In other posts, I’ve written about the distinction between “the work” and “the workplace.” I want to dig a little more into what makes workplaces unique.

It is not naïve to seek autonomy, mastery, and purpose at work. Given how much of our lives we spend working, it would be strange not to.

The seductive mistake, however, is to assume that companies function like families, guilds, unions, friendships, expeditions, sports teams, congregations, neighborhoods, or communities. Humans organize themselves in all sorts of interesting ways, each with its own purpose, expectations, obligations, and peculiarities.

Institutional theorists Roger Friedland and Robert Alford argued that different institutions have different “central logics.” A family, company, profession, and community may overlap, but they demands different things (from us), and make different promises in return.

Companies are unusual.

They can provide us with meaningful work, friendships, belonging, opportunities to grow, and, importantly, money to put food on the table. But they also combine several features that complicate our relationships within them.

• Our livelihood, career, and reputation are at stake. So high stakes.

• Relationships are temporary, uncertain, and often not of our choosing.

• Power and influence (and protection) are unevenly distributed.

• Performance and accountability are open to competing interpretations.

• We can survive without resolving a conflict (e.g., “lay low” or “wait it out”).

• What protects an individual may harm the team.

• Leaving is possible, but rarely without cost.

This is why companies are not especially well suited to providing unconditional support, lasting membership, or the kind of mutual obligation and reciprocity we might reasonably expect from other relationships. When we imagine companies as families (or teams, pirate ships, professional communities, or congregations), we transpose expectations the institution simply cannot reliably fulfill.

Denise Rousseau calls this a psychological contract, which she defines as “an individual’s beliefs regarding the terms and conditions of a reciprocal exchange agreement.” Because the contract exists in the mind of the employee, it can feel binding even when unwritten or not shared by the organization. A breach can therefore feel like betrayal even when no formal promise was broken.

Well-meaning HR teams and leaders may work tirelessly to create environments of inclusion, engagement, belonging, and psychological safety. That work matters. It can make companies far more humane. The care and relationships can be real. But none of this makes the bargain unconditional. It remains shaped by economics, incentives, unequal power, and the ever-present possibility of exit. The more a company feels like a community, the easier it is to forget the limits of the institution underneath it.

There is a lot of talk about agency, and especially being “high agency.” That may help us improvise, and roll with the punches. But it does not magically change the calculus of company work. The imbalance of power and the cost of leaving remain. The relationship is still conditional.

We also place expectations on ourselves that we cannot reliably fulfill. We can come to see our inability to resolve contradictions built into the institution as a kind of self-betrayal. When the bargain breaks down, the sense of betrayal can turn inward as a kind of self-gaslighting. We beat ourselves up for “caring too much,” staying too long, failing to speak up, failing to leave.

This is where healthy ambivalence helps. We can let the work and the relationships matter without expecting the institution (and the humans representing the institution in various ways) to promise more than it can.

Simone de Beauvoir wrote about the ambiguity of being both free and constrained.

Let us try to assume our fundamental ambiguity. It is in the knowledge of the genuine conditions of our life that we must draw our strength to live and our reason for acting.

We can act, but we always do so within conditions we did not choose or control. Something similar is true here. Work can be genuinely meaningful, and the workplace can remain a conditional bargain. Neither cancels the other out.

Most of us will work for decades. Across that time, we may have several “work lives.” The work itself may follow different arcs and change as we do. We may spend only a few years at a particular workplace, or perhaps a decade. Keeping that difference in view creates a certain freedom, even some levity. A workplace can matter deeply without becoming the whole story. It is one place where the work happens for a while.

And some lighter material… after The Standard Setter, I wanted to follow up with another curious character….

The Case-by-Caser

May I present to you: “The Case-by-Caser.” By repeatedly asking for specifics, they can make any problem not their fault, and any systemic issue magically disappear.

The tactic here is very simple. No matter the problem, keep pushing people to give “specific examples” until you’ve narrowed and flattened the thing down to something innocuous that can ideally be pinned on an individual, or a small group of individuals, or a concrete next step.

The Case-by-Caser makes it very clear that no systemic issues, or anything uncomfortable or messy, will be tolerated. The burden of proof falls completely on the reporter. Anything that doesn’t comport with some arbitrary (and safe) version of specificity will be summarily chastised and challenged.

The beauty of the CBC’s tactic is that, on the surface, specific observations are valuable (especially if you stick with observations and don’t label them). They are salient. They help people grapple with what they are experiencing. The right tactic is to surface broad/diverse observations, and then grapple with sense-making together. Act through “the problem.”

A problem that has stuck around is most certainly impervious to simply reporting a “specific example.” You have to foster a situation where the team can move beyond sweeping statements to something more “tangible,” but that isn’t a switch you just turn on/off.

So asking passes the sniff test.

With the CBC....

The problem is in the complete lack of coherence in their approach.

The definition of specific is completely arbitrary. When it comes to a vague strategy and someone calling that out, they are “uncomfortable with uncertainty!” “Why do you need all of these details?” Yet when it comes to a problem, somehow the tables are turned and you need to define the specifics. Strategy vibes are fine. Problem vibes are not fine.

Why is this effective? No one ever got fired for asking for specific examples. Plenty of people have gotten fired for trying to talk about systemic, messy problems.

So the CBC has a 100% self-sealing tactic. There’s a hard-to-parse undercurrent of fear and doubt. “What’s a specific example?” How do you answer that?

Whenever there’s a problem, the Case-by-Caser confidently points to the fact that they’ve addressed “all specific problems,” and they have a self-sealing argument to their boss.

Is there any way to beat the CBC at their own game? Probably not, especially if they have more positional power than you do. They can always ask for another example.

Stop trying to win the argument. Ask yourself what intervention you can actually make that has some chance of changing the system. You don’t necessarily need the CBC to agree with your diagnosis to start acting on what you’re seeing. Sometimes the most expensive mistake is believing you need permission from the person denying the pattern before you can do anything about the pattern.

May I present to you: The Standard-Setter. The Standard-Setter’s superpower is failing upward while confusing pressure with motivation. They believe everyone has another gear, and that their job is to force people into it.

The Standard-Setter operates from a simple belief: that by merely telling you to “make the impossible possible,” and by “setting extremely high standards,” they can somehow magically make you more effective.

But they often have no idea what they’re actually talking about. They don’t demonstrate the thing by example, or their example comes from a completely unrelated domain. What they have internalized, deeply, is the identity of a high-performance leader: someone who demands excellence, refuses excuses, and gets “more” out of people.

What they often dismiss is the fact that you could get the same results w/o driving everyone crazy and wasting energy dealing with them.

Statistically, it is probably safe to say that we are all “capable of more” under certain contexts, for certain periods of time, and with varying second-order costs. So the Standard-Setter will inevitably accumulate plenty of evidence that what they are doing “works.” Someone rises to the occasion. The mistake is treating those observations as evidence of impact rather than simply evidence that people can respond to pressure.

The Standard-Setter isn’t like a trusted coach who has worked 1:1 with LOTS of people over the years playing a very specific game. No, the Standard-Setter sort of naively adopts the worldview that pushing people harder is itself a form of leadership. And then, oddly, they expect some of the loyalty and trust we might give a great coach without having earned it through expertise, attention, shared experience, or demonstrated understanding.

The Standard-Setter also tends to hallucinate that they have somehow figured out how to motivate *everyone* with their approach, missing that a lot of people don’t respond to pressure, exhortation, or someone else’s intensity. They respond to clarity, trust, autonomy, meaning, mastery, good feedback, and the feeling that their effort is actually connected to something worthwhile.

A great coach or teacher knows this. They don’t try to motivate people the same way.

The Standard-Setter sounds like a broken meme, cajoling everyone with the same platitudes.

Standard-Setters can also make remarkable upward progress in organizations because their style is extremely legible from above. They look ambitious. They can point to moments when teams stretched and “won.” The costs, meanwhile, are often dispersed downward: exhaustion, resentment, churn, and the energy required to manage the Standard-Setter themselves.

If there is one solace, eventually the Standard-Setter has to show what THEY can do, or meet their match in someone with enough influence that they CAN’T bully. The trick only works as long as the Standard-Setter can keep converting positional power into apparent performance.

In this post, I’m going to share ten career traps thoughtful people often find themselves ensnared in.

I’ve found myself in every one of these traps over the years, so I speak from experience. I’ve also watched some of the most thoughtful (and experienced) people I know fall into them, even when they “knew better.” Or perhaps partly because they knew better. As sports coaches sometimes say, we go where we’re looking.

Every trap below begins with a strength or virtue. Curiosity. Empathy. Courage. Helpfulness. Sensitivity. Intellectual honesty. A willingness to question. A desire to make things better. The ability to see patterns other people miss.

What makes these traps interesting is that they usually start innocently enough. You get pulled in through some combination of your strengths, your weaknesses, and your strengths becoming weaknesses.

It isn’t that we have no agency. We do. The hard part is figuring out where the inflection point actually is. In hindsight, it can look obvious. In the moment, it often isn’t.

I’ve recently learned how much patterns, and naming the traps, can help. The trick is figuring out your tripwire: the saveable moment. You have to rehearse it before you’re in the thick of it. Find a friend to help. Plaster your wall with reminders (just off camera).

The Volunteer Trap

When your radar for problems renders you illegible.

aka “What did we hire them to do again?”

You sense incoherence, and you’re eager to help, so you self-volunteer to help out. Or you’re invited to help by people who sense your willingness to help. Many of these problems are not officially part of the mental model for your job (though they may obliquely block your job). They’re lingering problems, and there’s a reason they’ve stuck around. But your optimism and general belief in humanity, seduces you into believing this time it could be different. People are gracious at first. They are relieved, and even invite you in. Then you’re “doing things” people don’t generally understand and can’t explain.

Gradually, you become the figurehead for problems you’ve waded into, and when layoffs come, people “don’t get what you do.”

Saveable moment: You notice you’re stepping into a problem because you can see it and care about it, but it’s not clear who else is willing to own it, support it, or have your back when it gets messy.

See also:

• The Spokesperson Trap. Stepping into a messy, orphaned problem can make you the person associated with it. “What’s going on with X?”

• The Canary Trap. Taking responsibility for problems others avoid can pull more and more of your attention toward what is broken. People might actually use you as an early-detection system.

The Curiosity Trap

“We Want a Fresh Perspective, Just Not That Fresh.”

People are curious. They seek you out. They tell other people that you’ve been chatting “So I was talking to Cutler about this…”

But they’re only pulling out the parts they already agree with and scratch the curiosity itch. They fish for the actionable bits, can’t find them because you never really had a reference point, and you get pegged as “all talk” even if you were offering solutions. Alas, they didn’t interpret them as solutions because they didn’t match their version of solutions (to you, they were).

Someone who was validating you with their curiosity is now turning on you, and the message that “leader doesn’t like X” starts to spread. Your other relationships suffer. What drew them in—the outside voice, the person offering something new—ultimately becomes a source of threat.

Saveable moment: Someone keeps coming back for your perspective, but you notice they are mostly extracting the parts that fit what they already believe.

See also:

• The Thinking in Public Trap. People may seek an outside perspective, then recoil when that perspective comes with visible complexity and unresolved edges.

• The “Let’s Admit We Don’t Know” Trap. A fresh perspective often exposes uncertainty that people were hoping the fresh perspective would resolve. “I thought you could fix this, and now you’re asking questions?!”

The Defender Trap

When you jump to defend, but end up becoming the focus of the fight.

aka “Why are they standing up for people who are obviously wrong?”

You see someone, or a group of people, being ostracized or typecast, and it triggers a need to defend, push back, and come to the rescue. Voice the other side. Challenge bias. You hold on to the idea that it’s a misunderstanding and try to win people over with more information. That comes across as naive and difficult to parse (“I don’t get why they’re standing up for the people who are OBVIOUSLY to blame!”). In the version of the world the antagonizers inhabit, this is a sign of weakness. You become a sponge for people’s anxiety from all sides. Lose sleep. Start being a jerk yourself. It becomes us vs. them, you attract allies who are on people’s shit list, and you’re “so negative.” By that point, you’re also part of the problem because you have trouble offering generous interpretations of people’s behavior. What started as a misunderstanding, is too far gone.

Saveable moment: You notice yourself moving from “I want to make sure this person is being interpreted fairly” to “I need to prove these people are wrong.”

See also:

• The Spokesperson Trap. Defending someone else’s position can make you the most visible carrier of it.

• The Canary Trap. Staying inside a conflict long enough can turn hyper-vigilance into a way of operating.

The “Let’s Admit We Don’t Know” Trap

When your comfort with uncertainty gets mistaken for reluctance, perfectionism, or lack of conviction.

aka “Why are they making this so complicated?”

It’s OK to be uncertain. What really gets you is when people say it’s figured out, and it isn’t. So you ask questions, and seemingly innocent questions challenge the status quo. You get branded a perfectionist, a skeptic. All theory. Not realistic. Once the criticisms or “constructive advice” starts rolling in, you sure START sounding like a perfectionist. You defend your pragmatism with altruism, and your bias for action with apparent reticence. You’re spending most of time trying to defend your feedback and questions vs. actually getting answers. The more defensive you get, the more you’re actually part of the incoherence. What started as a leader choosing to gloss over the inconvenient parts of the story, becomes an all-out assault on your reality and your dignity.

Saveable moment: You have asked the question once or twice, and instead of answering it, people start giving you feedback about how you are asking it.

See also:

• The Right But Early Trap. Admitting uncertainty often means noticing weak signals before others are ready to take them seriously.

• The Thinking in Public Trap. Showing what remains unresolved can be mistaken for not having a point of view.

The “Thinking in Public” Trap

When showing your reasoning process gets mistaken for not having a point of view.

aka “Can we just get to the answer?”
aka “Do they ever just land the plane?”

You break things apart and rebuild them. You take joy in refining ideas with holding your ideas loosely and being willing to crush them. You’re comfortable with holding multiple competing ideas at once. That produces stellar results, especially when you’re able to control your groove and audience, until it becomes your shtick. Over time, you’re known for going deep, and seeing both sides, not driving action. Your thorough, thoughtful, and heartfelt narratives get compared against simple ones. Holding multiple perspectives reads as lack of conviction, and “how can we make this actionable?” follows you even when you self-identify as having conviction and having a bias for action.

As with other loops, soon you’re investing a ton of energy defending yourself from the typecasting, over-compensating, and suddenly you’ve gone from “strategic”, to “they go way too deep all the time,” to “they seem to be selling all the time.” You become self-conscious thinking in public or sharing your mad-scientist vibe.

Saveable moment: You notice that people are reacting to the shape of your thinking more than the substance of it.

See also:

• The “Let’s Admit We Don’t Know” Trap. Showing the unfinished parts of your reasoning (even if it is super fun for you, and some people are fans) can look like indecision.

• The Curiosity Trap. People may enjoy your perspective until the complexity behind it stops being easy to consume.

The Right But Early Trap

aka “Why are they making such a big deal out of this?”

You pick it up earlier, and it bothers you earlier. You’re strong at signal detection and taxed by attention gating. To you, calling it out is the solution. To them, “it’s easy to poke holes in things.” You raise a weak signal, and people react as if you’d asked to change the strategy. It accumulates, you get typecast, and your restraint cracks. Right != right now. And with nowhere for the signal to go, you keep pushing.

Saveable moment: You have raised the weak signal, people have heard it, and nothing is happening.

See also:

• The Canary Trap. Detecting weak signals over and over can leave you scanning for the next one.

• The “Let’s Admit We Don’t Know” Trap. Sometimes the signal you are trying to protect is simply that the situation is less settled than everyone says.

The Spokesperson Trap

aka “Everyone told me this privately.”

People share their tension with you, and you see a pattern across them. But then the question is: who’s voicing this? You end up voicing it, even when you try to create a forum for them. Then even the people who shared their concerns back away. They’re fine voicing it to you, just not being associated with you. It becomes your burden. And then suddenly your call-to-arms. This is the worst of all worlds in a sense. You fail at helping others have a voice AND begin to jeopardize the cause yourself. And if you taunt the elephant in the room, suddenly you’re wrapped up as a spokesperson for “Something no one else is describing.”

Saveable moment: Someone says something important to you privately, and you immediately start thinking about how to surface it.

See also:

• The Defender Trap. Once the issue gets attached to you, criticism of the issue can become criticism of you.

• The Volunteer Trap. Many spokespersons started out as people trying to help with a problem no one else owned.

The Workplace Identity Trap

When your experience in one workplace starts to feel like a verdict on your entire professional identity.

aka “Why do I do this anyway?”

Instead of identifying as a designer, you identify as “design at X.” A manager’s validation means “I did a good job.” A leader listening means “they believe in me.” When you fail there, it’s an identity threat. Even small comments send you into a spiral instead of a “shrug, oh well.” As your insecurity grows, so does the need to be validated by THIS environment, THESE people, and THE culture. It is a drift as old as time. What should be seen as a healthy arc across many companies over decades becomes a verdict on your abilities. Meanwhile, no one will be working in this company in 3-5 years.

Saveable moment: A piece of workplace feedback starts changing how you feel about the work itself.

See also:

• The “Game Is All I Know” Trap. The more local validation matters, the more tempting it becomes to reshape yourself around what the environment rewards.

• The Canary Trap. Enough cycles of conflict and exhaustion can turn “this place is hard for me” into “maybe I’m the problem.”

The “Game Is All I Know” Trap

aka “Wait, which part of this is actually me?”

aka “I don’t know how not to do this anymore.”

You learn to adapt, to play the game, at great expense, and you get good at it. You get rewarded for it. You’re playing the game with pride and dignity. It goes well for you, which is the problem. Your ability to mask effortlessly hides the fact that your core identity and life-force is hurting. You rise, showing your cards gets riskier, and you get more guarded at work and at home. Trapped by your own accommodations, until accommodation becomes your identity. This is a cousin of the work identity vs. workplace identity puzzle mentioned above, but it is extreme because your workplace identity isn’t even a playful dance around “the rules.” It has seeped into your identity all the way through.

Saveable moment: Playing the game gets easy.

See also:

• The Workplace Identity Trap. Adaptation gets dangerous when succeeding in one environment starts defining who you are.

• The Thinking in Public Trap. The better you get at presenting the acceptable version of your thinking, the harder it can become to show the messier version at all.

The Canary Trap

The more sensitive you are to what’s wrong, the more likely you are to become consumed by detecting it.

aka “The canary always dies.”

It starts off great. Full force, a lot of attention, a bit of a platform. The more you notice, the less coherent the company becomes, and the less coherent the company becomes, the more you notice. The more you fight. Flame up, flame out. On to the next company.

Saveable moment: You realize you are spending more time tracking the company’s incoherence than doing the work you came there to do.

See also:

• The Right But Early Trap. Being early over and over can train you to stay on alert for what everyone else is missing.

• The Workplace Identity Trap. Repeated flameouts can make an environmental pattern feel like a personal verdict.

Conclusion

The point isn’t to avoid all of these traps. You probably won’t.

The point is to notice them early enough that you still have your wits about you.

Before you’re exhausted. Before you’re defending your identity. Before every interaction feels loaded. Before you’ve spent so much cognitive energy on the situation that perspective itself starts to disappear.

That’s the saveable moment.

When you can still ask:

• What is actually happening here?

• What am I trying to protect?

• What happens if I do nothing?

• What happens if I step back?

Once the trap has fully closed, those questions get much harder to answer. You’re inside the thing, spending energy just trying to make sense of it.

So the work is not becoming less thoughtful, less sensitive, less helpful, less curious, or less willing to care. It is learning to recognize the moment when those strengths are starting to pull you somewhere you may not actually want to go.

Customer call → synthesize insights → decide what to do about it → do it → ship it → measure impact

Pre-AI, this would involve 1) reading/reviewing a call transcript and your notes, 2) synthesizing that into a set of takeaways either in a document or presented somehow, 3) considering a set of options by talking about those options, writing about them, and narrowing those down to some tangible actions/investments, 4) taking those actions (e.g., building something) using whatever collaborating techniques and “pointers to the work” (e.g., tickets), 5) shipping it, and 6) ideally measuring impact by going into an analytics tool, watching feedback channels, reaching out to customers, etc.

At each stage, you were, in effect, asked to process, think about, discuss, reshape, restate, copy-paste, “migrate” information down the line. There might have been meetings, routines, gates, swapping between tools. With only N hours in the week, you were constantly kicking yourself for not being as collaborative, inclusive, comprehensive, or rigorous/disciplined enough. Yes, this was a pain in the ass, but it was also “positive friction.” Each of these information migration moments was an opportunity, in HCI parlance, to apply a forcing function: to focus conscious attention on something, to avoid automating your thinking, to pay attention, discern, say “no” or “what if.”

The friction was judgement and discretion’s friend. We tend to remember the crappy parts: the meeting prep, the “I have a document with 35 feedback themes and a Miro board. Where the hell do I put this?” We remember that sense of never doing enough.

But it was also valuable.

New Reality (Unless You Push Back)

Say you wanted to just automate everything now with AI. You could effectively:

1. Do your expenses during the call. Just ask the occasional question.

2. Take that three-thousand-word transcript and ask an LLM to “summarize this into key pain points” (the average call can generate dozens if not hundreds of these)

3. Plop those pain points into a “skill” that looks at the current code, maybe even checks analytics, and writes up “specs”

4. Run the “Strategy alignment prioritization” skill (alongside your secret “what I need for my promotion” skill) and sequence the bets

5. Shuttle that into some kind of harness to build the thing

6. Have some other agents review the PR

7. Ship it…

8. Half-heartedly verify the features work, but hey, Playwrite has been doing a good job lately, so automate that

9. Trigger a downstream analysis skill that asks if it “worked”

10. And you could effectively do dozens of these calls, or just monitor customer “voice” channels and not even show up. Promotion! Bingo! What a time to be alive!

Notice all the opportunities for thinking, pruning, shaping, designing, judgement-applying that are lost. AND, noticing all the drudgery that is lost. Noticing how very quickly nothing you are looking at is “real” (original, atomic, as written by a human). You have tens of thousands of words of original feedback, yes, but very quickly all you have is pointers (if you’re lucky and/or smart). Yes, the code is real. That’s why everyone has the utmost confidence in that step. But how did the code get that way? Sure, if it is all obvious slop, that’s one problem. But what if it isn’t? That’s almost worse.

Prior to AI, you might focus on one product area, one opportunity, and methodically work through it. You did the same calls, but those calls didn’t come with hundreds of things you could/can do. The blast radius of any particular thing you were working on was narrower. Very quickly, you amassed more feedback than you could meaningfully act on, support, market, communicate to customers, and keep in your own head.

Now you can go from five customer calls (~8000 words) to 80 “opportunities” to 80 “tickets,” 400 “tasks,” and hundreds of stuff to show customers, to hundreds of bits of feedback on those things, to 50 “feature adoption dashboards” (hidden behind MCP obviously).... Not to mention all the internal call recordings, status updates, PRs, stuff, fluff, slop, 1000 interface variations, etc. And after doing ALL of that, your customer is going to say, “I can’t keep track of everything you’re doing.”

All About Relationships

I recently did an activity where I looked at hundreds of artifacts at my current gig and looked at the “job” each one of them played. Many artifacts contain LOTS of atomic bits of job-doing information.

• Signals: What is observed, recorded, written, or measured

• Insights / Interpretation: What we infer from those signals

• Models / Hypotheses: How we believe things work and connect, including assumptions

• Options: Possible paths, configurations, or interventions

• Choice: The selected option and resulting commitment

• Intent: The future state or direction being pursued

• Actions: What actors do to advance the intent

• Mechanical Change: The specific thing that is modified

• Effects / Impact: The resulting change

• Actors: Who observes, interprets, chooses, acts, or is affected

• Materials: Systems, tools, technologies, artifacts, and infrastructure

• Constraints: Conditions that limit, shape, or enable the system

Whenever I see a simplistic “SDLC” diagram, I laugh because real knowledge work is much more than a couple stages. The ontology is much richer.

Yes, there might be a sort of physics in the flow (e.g., observe, orient, decide, act), but imagine everything above as a graph. Every edge in that graph is a transition point, even when we lump together ideas. A PRD might contain signals, insights, choices, etc., and itself serve as a model, but you are really dealing with a discrete mix of more enduring/lasting ideas and more transient ideas.

Of course, the problem was there pre-AI as well. Very few people thought in this level of detail because “it was too complicated” and “you do this naturally when you’re solving problems.” But the active participation was key.

We tend to think of the nouns. But it is the edges where the work happens. Pre-AI this happened whether we liked it or not. Now…we have excuses.

At work, I was trying to untangle an initiative recently. What I observed was as follows:

• TONS of AI-generated summaries on top of summaries

• Lots of surface area in an attempt to be customer-centric. We had a lot of change in the mix, probably far outpacing the customer’s ability to absorb or provide feedback on that change

• AI is very lazy at telling the difference between things (e.g., signals and insights, intent and actions, etc.)

Unless we are very deliberate, we end up with Linear tickets being a dumping ground for synthesized feedback (AI) and code review (AI), and you couldn’t help but be trigger-happy and just ship, ship, ship.

Principles

Here are some rough principles I’m using to tackle this problem:

• Keep original feedback atomic, and preserve its path back to the source. Transcripts and feedback are records of the original context and the start of a loop. Whenever we break that material into smaller items, such as an issue, observation, theme, or action, each item should link back to the original transcript or feedback. That traceability lets us check what was said and recover the surrounding context. If you can’t link back to something real, stop doing it.

• Preserve both atomicity and coherence. Product artifacts often mash together distinct ideas, such as Intent, Options, and Change, into a single issue or document. AI needs those ideas represented as separate, atomic pieces while keeping their relationships and context intact. Atomicity without coherence leaves fragments; coherence without atomicity leaves ideas mashed together. We need both. Miss either and we’re out.

• Add positive friction when content is copied and recontextualized. At each transformation, make it clear what is being carried forward, what changed, and why. The friction belongs at that moment: enough to make the transformation intentional and keep the source links intact.

• Separate the underlying job from its presentation to different audiences. Different audiences may need different framing, but that doesn’t change the underlying job or intent. Keep the source clear, label each audience-specific version as a variation, and link it back to the source.

• Label AI-generated content by degree of AI contribution. Make the label visible on every AI-generated artifact, with a variation such as “100% AI-generated,” “HITL curated,” or “AI used for editing/grammar.” Keep the label with the content when it is copied or recontextualized.

• Avoid compounding AI flattening. The biggest anti-pattern is AI flattening layered on AI flattening, with no thread back to reality. Each round can strip away more of the original context while making the resulting abstraction look authoritative. Summaries are most useful when they get close to a specific problem, opportunity, or action, and remain linked to the source material.

• Don’t let AI capability turn a tool into a dumping ground. The fact that a tool can process information with AI doesn’t mean everything should be put there and sorted out later. Give each item a clear purpose, and keep it connected to its source and context.

• Treat code analysis as one source of evidence. It can be very helpful to see “what the code says,” but seeing X in the code doesn’t establish that customers experience X, that we’re solving the right problem, or why the code came to be that way. Don’t let the concreteness of code analysis create confidence about the broader product reality that the code alone can’t support.

• Use Linear for bounded team intent. An issue should make clear whose need it represents, why it matters, what problem we’re addressing, and enough of how we’ll address it to make the work verifiable by people or machines. It should capture intent, not serve as a transcript repository.

• Go for depth over breadth in product improvements. If everything changes at once, customers can become confused, their feedback can become scattered and nonspecific, and the data can conflate the effects of different changes. Focus improvements enough that customers can respond clearly and we can learn what made a difference. Being trigger-happy with new features puts us at risk of taking on too much breadth.

• Recognize Linear has two jobs. One is team-visible product and feedback intent: the problem or opportunity we’ve decided to pursue. The other is personal planning and decomposition as work unfolds. Both help people track progress and preserve context, but personal breakdowns can create noise for the team and may fit better in personal notes or another workspace.

Hope sharing this helps you think about how you work!

As a last word, I’d leave with you the question:

Where was positive friction a good thing? And how can you embrace new technologies without losing the good things?

Personality and “strength” tests at work have never sat right with me. Sure, they became fodder for some fun discussions and had that occasional “wow, how right!” horoscope vibe. But they rarely seemed to help with day-to-day work, and they never replaced the hard work of truly getting to know people, including how they interact with others and how different contexts bring out different versions of them.

When building a team, we tend to adopt the sports-team metaphor: find the right people for the right positions. Or, slightly more generously, ask, “How do we bring out the best in this person?” or “How can we get these people with different strengths and weaknesses to do great things?”

Though phrases like “culture fit” acknowledge, at least in theory, that someone might be highly successful in one culture and struggle in another, in practice they often downplay the quirks and idiosyncrasies of the culture itself. The culture gets translated into a set of seemingly context-free strengths: decisive, collaborative, high ownership, low ego, fast-moving, intellectually honest. In short: the way we already are becomes “good,” and then we attach big, admirable words to it. The focus then shifts to figuring out how the individual will “fit.”

Even when we know better, we often take the easy way out when describing other individuals. “Oh, they’re a strategist.” “They’re a people person.” “They have high conscientiousness.” “They’re low in agreeableness.” “They’re very analytical.” The label is convenient because it compresses a lot of complexity into something easy to remember and repeat.

This post is about “types,” cultural norms, and why we should focus much more on how people show up in different contexts than on forming theories around the buckets people fall into.

Types And Other Boxes (At Work)

Most personality frameworks attempt to describe relatively stable tendencies. But what you actually care about at work is more like: “What will this person do in this specific kind of situation, with these incentives, norms, power dynamics, and constraints?”

“What type of person are you?” is the wrong, but convenient, question. A better, but harder, question is: “Under what conditions do you behave differently, and what distinctions are you making that cause you to switch modes?”

Take two people who rank high in conscientiousness and low in neuroticism on the Big Five personality test. They are tasked with figuring out the meeting cadence for the company.

One says, “We’re wasting too much time coordinating. Let’s set a standard cadence, put it on everyone’s calendar, and adjust it in a month.”

The other says, “A company-wide meeting cadence is hard to undo and affects everyone differently. Let’s understand where coordination is actually breaking down before imposing a new default.”

Both responses are perfectly compatible with being conscientious and emotionally stable. They have the same personality traits, based on the model. What differs is how they interpret the decision. How reversible is it? What evidence is required? Who has the legitimate authority to decide? What counts as an acceptable experiment?

High conscientiousness and high agreeableness could plausibly produce either:

“This is chaotic. We need a clear daily cadence, but we should involve everyone in shaping it so people feel heard and can commit to it.”

or:

“A recurring meeting is a significant organizational commitment. We should diagnose the problem carefully and get broad consent before imposing a new default.”

Same broad traits. Very different behavior.

The discussion here extends beyond personality. The same issue shows up in strengths frameworks, skills frameworks, leadership competencies, communication styles, and other models, frameworks, and systems that try to describe people in relatively stable terms. Knowing that someone is “strategic,” “collaborative,” “decisive,” or strong at influencing tells you something. But it still leaves open the harder question: under what conditions does that strength show up, fail to show up, become overused, or look very different?

And don’t get me started on “growth mindset,” a phrase I’ve heard bandied about by people with remarkably fixed ideas about what good thinking, good behavior, and good leadership are supposed to look like.

A Personal Example

I’ll use myself as an example. Over the years, I’ve taken everything from Myers-Briggs and DiSC to StrengthsFinder and more empirically grounded trait assessments like the Big Five.

In most situations, I put a high premium on co-designing the solution with other people. I like getting perspectives on the table, making sure people feel some ownership, and then moving. I’m not especially interested in consensus for its own sake; I still have a bias toward action. But I generally want the action to feel like something we shaped together rather than something that was simply imposed.

In a different kind of situation, though, I can change modes very quickly.

If I perceive someone as abusing their power or influence, generalizing about a group of people, dismissing someone’s experience, or treating another person as if their perspective doesn’t matter, I jump into defender mode. I become much more direct, much more willing to confront, and much less concerned with preserving harmony or being seen as agreeable.

So from the outside, those two versions of me might look inconsistent: collaborative and accommodating in one setting, aggressive and confrontational in another. If someone assigned me Big Five personality traits based purely on my behavior in those two situations, they might come away with very different profiles. But internally, the behavior is connected. In the first case, I’m trying to create shared ownership. In the second, I’m reacting to what I perceive as someone using power in a way that makes shared ownership impossible.

In fact, I even received feedback that I was “polarizing.” This stung. After a lot of soul-searching, I realized that I was showing up very differently in certain situations. The feedback wasn’t wrong: my behavior could be polarizing. But it was also internally consistent. Certain situations pushed me into a much more forceful, aggressive defender mode, while in others I was highly collaborative and accommodating.

“That’s Just How Things Work Around Here”

There’s a strange asymmetry here.

Organizational norms are themselves the accumulated byproduct, or sediment, of people’s behavior. Behavior shapes the rivulets and channels water passes through. But once those behaviors become established, they often get depersonalized into “that’s just how things work around here.” The norms become part of the environment, seemingly detached from any particular individual.

Anyone who benefits from those norms, or contributes to them unconsciously, gets to point to the channels as if they were simply features of the landscape. “That’s the way water flows!”

The person adapting to that environment, however, is treated very individually. They need to read the room, adjust their style, develop better judgment, learn the context-appropriate form of “executive presence,” or learn how to operate in the culture. The behavior that created the context becomes invisible, while the behavior required to adapt to it becomes a personal competency and burden.

Various academic and scientific traditions address this head-on:

• Institutionalization (Peter Berger, Thomas Luckmann, Lynne Zucker): repeated human practices can eventually become taken-for-granted features of social reality.

• Organizational culture (Edgar Schein, Joanne Martin): assumptions and ways of working can become so embedded that people experience them as simply “how things are done.”

• Privilege and social dominance (Jim Sidanius, Felicia Pratto): dominant groups and their ways of operating can become treated as the unmarked, “normal” case.

The burden does not fall evenly. It often falls most heavily on people who are underrepresented twice over: underrepresented in the immediate environment they are trying to navigate and underrepresented more broadly in the profession or industry. They may have fewer people around them whose default ways of speaking, leading, disagreeing, or making decisions resemble their own. So they are asked to do more of the adapting while having less power to define what counts as normal in the first place.

Courage!

The dynamic shows up when leaders throw around words like courage, discipline, “bias to action,” or ownership. These sound like relatively context-free virtues. Who wants to argue against any of those things?

But in practice, organizations don’t reward courage, discipline, action, or ownership in the abstract. They reward particular expressions of them. In one company, courage might mean challenging the CEO in a meeting. In another, the exact same behavior might register as disrespect, poor judgment, or an inability to read the room. In one company, ownership might mean making the decision and moving without waiting for permission. In another, doing that might get you feedback that you failed to bring people along.

Courage becomes speaking up in this particular way. Ownership becomes taking charge in this particular way. Discipline becomes persisting in this particular way. Bias to action becomes moving at this particular speed, with this particular amount of consultation.

If that particular expression maps neatly to how you already behave, it becomes very easy to experience the virtue and your preferred expression of the virtue as the same thing. Now someone new shows up. We think we are evaluating whether someone possesses a virtue, but we may actually be evaluating whether their expression of that virtue is culturally legible to us.

Conclusion

If we actually want to work better with people, it isn’t enough to understand personality, skills, or values in the abstract, yours or theirs. We have to understand how those things show up in distinct contexts and how the context we’re operating in changes what gets expressed.

It helps to understand your own or someone else’s general tendencies. But for team building, it may be even more useful to understand the contextual factors that bring out different versions of that core.

• What kinds of situations cause you or them to switch modes?

• What are you or they noticing in those situations that someone else might not notice?

• What assumptions are you or they making about risk, authority, fairness, reversibility, or responsibility?

The goal isn’t to abandon the idea that people have relatively stable tendencies. It’s to understand that those tendencies are expressed through context. Knowing your own or someone else’s “type” may tell you something about a center of gravity or area of inertia. Knowing the conditions that pull you or them away from that center may tell you much more about what it will actually be like to work together.

I recently started a new job, which means I’m spending a lot of my time trying to figure out how everything works: the product, the workflows, the history, the people, the artifacts, and all the context that everyone else has accumulated over time. So instead of trying to turn that experience into a neat argument, I thought I’d just write down what one of those days actually feels like.

8:30. Kid-school tornado is over. Second cup of coffee.

I’m building an elaborate “subway map” with stops, lines, and transfer points to wrap my head around a product. Certain stops can share lines, lines can loop, and transfer points let you “transfer” information, artifacts, handoffs, outputs, etc. I also have to account for the different humans and roles on these journeys, the signals that indicate whether different steps are working, and the various decision points.

Imagine a customer journey map turned into a subway system.

Why am I doing this? To make sense of things. To ponder what “end-to-end” actually looks like. To figure out how to pressure-test whether the whole experience fits together.

Will I keep it? Will anyone ever look at it? Who knows.

Customer call.

These sensemaking artifacts always outnumber tickets 8:1 in my world. Try to graft them onto tickets as labels or references, and it always feels disconnected. And one artifact is never enough. I always end up with journeys, a bunch of different map types, docs, atomic insights, metrics, one-pagers, goals, “Jobs,” and principles.

I take a couple of minutes to convert the subway map into a set of tables in Notion, on the off chance that I *do* want to do something with it. OK. That’s pretty effective. It would have taken me a while to set all of that up before. Notion’s AI makes it easy.

Must resist…

Must resist…

Step away from turning Notion into a subway system tool.

Meeting saves the day.

In addition to the subway map, I have a big list of tagged artifacts, lists of data sources and how they connect to the various subway stops, “screens,” and other access points (e.g., MCP, chat in other workbenches). Oh, and I have meeting transcripts, notes, Miro boards, videos of usability tests, and [insert 10 other artifact types].

And the product/experience itself, of course.

One puzzle, since I’m new at a job, is figuring out what is “still valid.”

Do these notes still make sense? Has the customer’s business changed? Is this an AI summary, or did someone actually fret over every word? Sometimes it’s easy to tell. Sometimes it isn’t. Is this definitive and pressure-tested? Does it represent any sort of convergence? Or did everyone glance over the doc and go about their day?

Anyone who has been around for a while probably has a ton of implicit knowledge about the situation floating around in their head. We have a couple of context-sharing sessions, with me furiously taking notes. Extremely helpful! The rest is me doing archaeology, trying to “connect the dots.”

I’m using AI the whole time, but for some reason nothing comes out great. I ask questions across all of these artifacts, and the answers are meh. They point me toward helpful things, but it’s very hard to tell which artifacts are “live” — continuously updated and reflecting current understanding — versus relics and cave drawings.

I’m worried about building shared understanding on top of eroded shared understanding. Or, worse, glossy, coherent-sounding, but ultimately wrong understanding.

I go back to building my subway artifact. There’s a tactile quality to it that I find calming. I can’t rush it.

I feel tempted to let Notion move it forward…

Tempted….

My willpower drops, and I pull the trigger:

“Oh hey, you have information about a lot of other workflows. Can you add the next couple of lines to my subway map?”

UNDO. UNDO. UNDO.

Meeting saves the day.

I can’t explain all the details, but I have to use PowerPoint for the first time in a lot of years. I used to work at a PowerPoint factory at an investment bank on the graveyard shift.

I don’t recognize PowerPoint anymore. I futz around with some AI features that don’t seem to work. I just want to paste without formatting.

Getting hungry.

Lunch (aka “Should I order Door Dash”).

I jump over to Linear and find that the tickets aren’t much good. I mean, they’re perfectly good for what they’re good for, but they aren’t designed for someone parachuting in with very little context. I think to myself, “this is how it should be.”

And then to the code. The code, at least, is definitive. It *is* what’s in production. So that’s helpful. If I want to know what this thing DOES and how it actually works, I’m not too far off.

I play around with AMP for a bit. It seems super powerful.

Some meetings.

More PowerPoint.

More subway system building.

Hear about the Miro acquisition. Miro is one of my favorite ways to work with people.

Some reflections….

You don’t appreciate the power of context until you’re new somewhere. But you ALSO appreciate the active nature of *creating* context through exploration instead of having it spoon-fed to you through a three-times-removed AI Slop Cascade.

I feel alive just trying to make sense of things.

At every step of the way, I find myself wishing someone had more time to work with me to forge even more powerful context.

It’s interesting: 10 minutes with AI can feel super-expansive in its breadth and its ability to answer questions, but it doesn’t have the same flavor as 10 minutes with a new coworker who has a ton of implicit context.

I can certainly see the power and potential of AI, and I enjoy using it in Notion. I haven’t really used AI inside a product like that before. But one thing sticks with me: I see a lot of Notion marketing around context, but I don’t think they’re hyping the act of context *creation and preservation* enough. I hear people talk about a context “battleground,” but it all feels vaguely reminiscent of how people used to talk about knowledge repositories, sources of truth, connected data, etc.

In other words: one part of the equation, and the death of many knowledge-management efforts.

I also catch myself, on a couple of occasions, fantasizing about “connecting” all of my thinking into a “brain.” But I’m quickly reminded that context starts eroding the second the day is over. Some things stay durable, but a lot of it changes subtly.

So that brain needs more information.

My brain is tired.

My book co-author sends me a text to check something out.

I do a walk-and-talk and have AI summarize my reply.

Shoot, I forget the “JUST SUMMARIZE” skill.

Redo.

End of day.

Have you ever felt like a complete weirdo in your company? Like there are only a handful of people who “get you”? Or have you gone to a conference and gotten a little confused because suddenly it seems like you have a lot more in common with everyone? Then you return to work, and somehow the weirdness comes back.

You may be less alone than you think.

One mistake we make is assuming that because an organization has certain rules, and because a handful of high-profile people seem to behave a certain way (and everyone seems to visibly go along with it), that there simply aren’t many people like us there.

Organizations can be more diverse in belief than they appear to be in behavior. What you see is not simply “what people here believe.” You’re seeing at least three things layered together:

1. Power-weighted norms, where some ways of behaving carry much greater institutional force and consequence than others.

2. How people adapt to that environment: what they say, suppress, emphasize, tolerate, and avoid in order to operate effectively within it.

3. The actual people underneath, who may hold a much wider range of beliefs, values, preferences, and ideas about how work should happen.

The key is that we mostly observe #2, attribute it to #3, and underestimate the influence of #1.

Leaders influence what gets rewarded, promoted, tolerated, and interpreted as “leadership material” and “good performance.” People adapt accordingly. That visible adaptation can then look like widespread agreement with the norms themselves.

This can create a reinforcement loop. People see others behaving according to the apparent norm and reasonably infer that those people believe in it. Their own adaptation then becomes another data point suggesting the same thing.

A relatively narrow set of norms can start to look like “what everyone here believes.”

But adaptation is not the same thing as endorsement.

People can have very different reasons and go to highly varying ends to play the part their work demands. They may be optimizing for belonging, safety, ambition, professionalism, coordination, or simply making it through the day. It is easy to see this as inauthentic, but the behavior is visible.

Leaders are adapting too: to expectations from above, peers, boards, investors, customers, and inherited ideas about what leadership is supposed to look like. The fact that someone is rewarded for performing a norm does not mean the norm reflects what they would naturally prefer, or even what they privately believe.

And there’s another factor: a lot of things need to fall into place for people at work to reveal certain beliefs.

• Some of it is luck: you happened to run into each other in the right setting.

• Some of it is time: enough time to create trust and dig into each other’s perspectives.

• And some of it is context: the right conversation, problem, relationship, or moment has to create an opening for that part of the person to actually show up.

Consider how many teams you’ve been on where it took quarters or even years to learn the why behind a certain quirk.

Social media has a similar norming effect. A relatively small, highly visible subset of people can make certain beliefs and behaviors appear far more common than they actually are. We observe what gets surfaced, mistake visibility for prevalence, and adjust accordingly. And some norms simply feel safer to perform than others.

LinkedIn is especially interesting here because we are still, in many ways, at work. People are visible to colleagues, managers, customers, future employers, and professional peers. So what gets expressed is shaped not only by what people believe, but by what feels professionally legible, rewarded, and low-risk.

That can make a narrow set of “safe” workplace beliefs look even more universal than they are.

In closing…

We encounter only a tiny sample of what the people around us think and believe. It is very easy to mistake what has been revealed to us for all that is there.

There may be far more “weirdos” around you than you realize.

For various reasons, it is normal to meet a couple of trusted people who really get you in your company. But you can’t let that convince you they’re the only ones.

Here’s an area where I am optimistic about AI (for now, at least).

So many companies flatten things to meet the needs of the person with 1) the most power, 2) the lowest appetite for detail/seeing real-world complexity, and/or 3) the least amount of time to actually work through details and nuance. For good (and less good) reasons, a CEO says, “This is too in the weeds. There’s no way we are doing this much!” That one statement will kick off one of a couple of things:

1. Investing huge amounts of time in flattening reality to get that report to the “right” level

2. Not investing that time and just winging it last minute. A reasonable survival mechanism.

Sometimes #1 is shouldered by lots of people in the org. And sometimes the burden is shouldered by just a few people. But it boils down to few/many people translating one version of reality to another, simplified, flattened version of reality. The recontextualization tax can be massive. So massive, in fact, that companies often slip into just winging it because no one in their right mind would want teams dedicating that much time to pretending things work one way in order to get “simple” overviews, “simple” cascades that “roll up perfectly,” etc.

If you’ve ever been in these high-level meetings, you understand that the desire for simplicity and “less” is most often a survival mechanism. It isn’t petty. It reflects a real cognitive load barrier. They are hoping to “manage by exception” instead of viewing everything.

Which brings me to AI.

An area AI excels in is doing the grunt work of recontextualization and exception hunting (provided, of course, it has the raw context and that context is reasonably up to date). Specifically, if you have teams working in highly effective, emergent, varied ways, but you really want to flatten that information more artfully into one format, AI is a good tool.

Say I have 10 stakeholders, and each stakeholder likes information “a certain way.” Pre-AI, I didn’t have many options. I could:

1. Try to persuade them all to look at information the same way

2. Make 10 versions of every presentation so that we could have effective meetings

This is why roadmapping tools almost never travel around orgs: you have so many stakeholders/audiences, and they each have different needs out of the roadmap, which means that it is far easier to just PPTX or Miro it up ad hoc every time.

Now, you can let a team work however they want to work and write a skill for each of those audience segments, tailored to how they understand product, think about roadmaps, care about roadmaps, etc. It doesn’t matter if you want to basically work in GitHub issues and Google Docs; you can pretend that you work in nice swimlane roadmaps with clear deliverables and an orderly flow. Or in a outcome-oriented “move the metric” kind of way. Or whatever fits your fancy, and your audiences way.

At the new job, we don’t want to force different clients to play by our rules. We also don’t want to hyper-customize Linear to become a super-abstracted, “respond to every client workflow” tool. Clients want things their way and they come first. We want things our way so we can make our clients awesome. So my first area of focus is figuring out how to put together skills that translate how we need to work to a language that is empowering for every client and doesn’t force them to use words/concepts that they don’t want to use.

AI is great at this.

I get all the high-end AI use cases, but a lot of the value is doing something you wanted to do forever but didn’t have enough hours in the day to make happen.

You don’t reinvent the wheel. Just do what you knew you should do, but couldn’t.

I can’t get this out of my head.

I was doing book research recently, and I learned about a “secret” group inside a company called the Right But Early Club. The basic idea of the Right But Early Club, if I understand correctly, is to serve as a support group (of sorts) for people who are “right” (we’ll discuss this more in a bit), but who tend to be “too early” when it comes to bringing up issues, calling out problems, offering solutions, etc.

Ever since I learned about this, it has been ringing in my ears. I’ve been telling everyone about it and have learned as much from people’s responses as I did in the original interview.

I think we can all relate, in some way, to the *feeling* of seeing something before other people see it, or being “right” before other people have arrived at the solution we believe is right. I’d divide that into two feelings and one decision: 1) seeing it, 2) it sticking in your head and permeating your thoughts, and 3) doing something or saying something. We’ve also been on the other side, being peppered by people who were confident they were right, knowing that technically they were (at least partially), but also knowing it was too early. To quote a VP friend, “I see plenty of things that I know will break, but I’m able to filter them out so I can focus on today’s problem.”

Signal detection and attention gating are both important cognitive abilities.

The puzzle is that many RBEC people (or more accurately people in the RBEC mode, because we can be many things in different contexts) are unusually strong at the first and taxed by the second. Organizations need the first while (many people working inside them) disproportionately rewarding the second. 

“Gripe Club!?”

When I describe the RBEC in other contexts and ask what people think, I often end up really triggering some people. “Those are the people who just bring problems!” “It is easy to be right about problems!” “Sure, but do they do something, or do they just sit in their support group complaining?”

I think this is important for a couple reasons. Primarily because how deep the tensions run.

First is that some people can’t seem to let go of the idea of rightness without reward or effectiveness. “They must be bad at communicating the solution!” The idea that there is a group of people in an org with the right answer who aren’t being rewarded or heard is immediately translated as an individual problem because the formula, based on certain world and work views, doesn’t compute. And let’s be honest, it is kind of threatening

Imagine if the CEO knew that there was a club internally filled with people who were correct at figuring out which projects would fail, which campaigns would flop, which hiring decisions would cause material harm, but that somehow no one listened? Pissed! But that explains the immediate rationalization and blame. It has to be the person, right? Or a group of people—the middle managers hiding good ideas.

On the RBE side, you experience very strong feelings around fairness, acknowledgement, and sensitivity to people glossing over disagreement. You may notice not only that something is wrong, but also who is being asked to absorb the cost of everyone else’s convenience. A quick “let’s move on” can feel less like pragmatism and more like a refusal to acknowledge what is actually happening. This can make it difficult to let an issue go, especially when the disagreement seems consequential or when the people with the least power are the ones expected to adapt.

Identity Clash

Notice the core identity clash here. It strikes all the way to how they view competence.

For the RBEC person, competence may feel like: I notice what others are missing. I trace consequences. I protect the system from preventable harm. I don’t look away just because the problem is inconvenient or premature. Their identity is partly bound up in foresight, coherence, vigilance, stewardship, and intellectual honesty.

For the Right Now person, competence may feel like: I know there are a thousand possible problems. My job is to decide what matters now, create movement, make tradeoffs, and not let hypothetical future concerns paralyze us. Their identity is partly bound up in agency, focus, pragmatism, prioritization, momentum, and responsibility for outcomes.

(Note: We can be both of these people depending on the context)

As I mentioned earlier around signal detection and attention gating, it isn’t that either identity is right or wrong, it is more that many organizations need both capacities but are structured to reward only one of them. They need people who can detect weak signals, trace consequences, and notice what others are missing, but they also need people who can filter, prioritize, and act without being overwhelmed by every possible future problem. The tension emerges when the organization treats one capacity as competence and the other as dysfunction, rather than recognizing that both are necessary at different moments.

“Perfectionists!”

There’s a big stereotype that people in the RBEC take “being right” very seriously vs. achieving some sort of outcome they care about. That happens, sure, but I think it happens less than people think.

Let’s differentiate a smug “we’re right before anyone will accept the truth” kind of vibe and a “Yeah, we accept there is probably a timing issue, and that acting too soon might be premature, and that we struggle with raising these issues when we’re working from pattern matching and signals that might not easily translate.”

At least in my experience, people who are right but early often have a mix of skill and experience in the domain and are strong in cognitive abilities like pattern recognition, systems thinking, and counterfactual reasoning (stemming from both their skill/experience and natural wiring).

They may also be particularly tuned to issues around justice, equity, or preventing pain, which heightens the emotional connection. They also don’t find it as easy to filter out noticing, and therefore know they bring things up prematurely. All to say that the smug vibe stereotype may reflect some situations—I’ve been on both sides of the “accept the truth” dynamic—but it doesn’t reflect all, or most situations in my experience. These are people with normal egos, normal career identities, etc. who, for various reasons, find themselves in the RBEC.

Being “Right”

Part of being right but early is acknowledging that mere “rightness” is not enough. Context dictates why something is right. Say there’s no senior leader with the express responsibility to improve X. Being right that X needs to be improved, or being right about how to improve X, doesn’t do us much good. Or say that a founder’s world view will kill any effort that feels like [something they don’t like]. Right Now is also relative to the other fires that are burning or questions begging for answers. No one sees it all. Companies often have elephants in the room commanding a ton of attention among leaders that no one knows about.

Right != Right Now OR Right Here.

I think what is fascinating about this is that you could have a company holding two truths: that it might not be taking advantage of or harnessing a certain population’s ability to be right earlier, AND that the people who are right earlier have their own set of issues and challenges that they are dealing with.

They may understand that they are not being as effective as they could, AND at the same time there are environmental factors limiting their effectiveness. They realize they can be right and wrong at the same time.

Community

What stands out most to me about the RBEC is the idea of people leaning into their quirks and situation and building community connections. To me, that shows a very high level of maturity in the situation. In most organizations, you are lucky enough to meet a small group of people who share your experience, who you can trust enough to admit being “right but early” as a mix of a shared experience and a shared challenge.

Whenever I go to conferences, I am blown away by how deep the conversations get. It is as if people haven’t had a single moment to “speak freely.”

A final thought. I shared a draft of this piece, and someone responded, “Oh, I’ve been in one of those, like a gripe club!” And I think that’s sort of the heart of my curiosity right now. “Gripe club” doesn’t do it justice, IMHO. That still diminishes the skills and strengths of the people involved—and the tension they are navigating.

Being right but early.

If any of this resonates, maybe you should set up a RBEC in your company?

Being Righter Together

Say you’ve done that…there might be a path forward for everyone. You still want and need a RBEC. But there could be other options…

The goal isn’t to create a company where the Right But Early people are listened to more because they are secretly right. That just creates another version of the same problem. The goal is to create conditions where people can be righter together.

There is a huge difference between bringing up something early and demanding that everyone act on it early. But many organizations don’t provide much space between those two things. You raise a concern, and suddenly the group is debating whether to stop the project. You point out a weak signal, and people respond as if you’ve recommended changing the strategy. You notice a pattern, and now you’re expected to prove the whole thing.

Under those conditions, bringing something up early becomes expensive for everyone involved.

What if, instead, there were ways to say, “I think I’m seeing something. I’m not sure yet. I don’t think we need to act on it right now, but I think it is worth noticing.” Now other people have something to participate in. They can add evidence, challenge the pattern, connect it to something they’re seeing, or help establish what would need to happen for the issue to become more important. The Right Now people don’t have to surrender the present, and the Right But Early people don’t have to pretend they didn’t notice the future.

This seems like an important organizational capability. Companies need ways to revisit things when new evidence arrives. Otherwise, the only choices are “do something about this now” or “stop talking about it,” which is a pretty terrible way to learn.

And I suspect this changes the RBE dynamic itself.

If people know there is somewhere for an early signal to go, they may feel less compelled to keep pushing it. They don’t have to win the argument today in order to prevent the observation from disappearing. At the same time, other people get to participate in making the observation better. Maybe the person is right. Maybe they’re directionally right but wrong about the mechanism. Maybe they’re right about the problem but wildly early about the timing. Maybe someone else sees a constraint they don’t see.

I was active on Twitter for a long time. I likely met many of you during that time. I recently collected 50 threads and 500 tweets (many with images) in this simple viewer.

Check it out for a blast from ye olde Twitter.

I also made this strategy prompts activity for teams hoping to get good strategy conversations going. Give it a try with your team! Let me know how it went.

Remember, a real strategy should contain propositions that challenge something we would otherwise take for granted. The real test is doing the activity in multiplayer mode.

Until next week.

Everyone is talking about “return on tokens.” Vendors love it (as long as the news is good). Companies that had long since given up on any sort of value architecture or on understanding the ROI of product work are suddenly chomping at the bit. Which makes sense: they haven’t been able to lay off enough people to cover their token budgets while still hitting their promises to the board.

It is hard to know what sits below the surface: a genuine effort to understand how AI might augment human capabilities, or some version of “how many people can we safely fire?” Or both?

It is as if the whole world suddenly shifted from a buffet model to buying each dumpling one at a time, and now everyone is freaking out about the ROI of each dumpling. But we’ve been here before. We’ve been here with hours. We’ve been here with capacity. And the problems have always been roughly the same:

1. Spending more time on the “I” side than the “R” side of ROI.

2. Myopically choosing shorter-term, easier-to-attribute use cases for the “R” side.

3. Gravitating toward whatever is easiest to measure. And tokens are very easy to measure.

This post takes a trip through hours and flow metrics, detours into an alternative way of thinking about ROI, and eventually gets to TOKENS—everyone’s favorite dumpling.

My goal is to show that the underlying measurement problems are remarkably universal. AI introduces some genuinely new twists, but in the end, it still hinges on a theory of value.

With these comparisons, ask yourself, “How might this relate to the current return on tokens question?”

Leave a Coffee Tip!

Hours

Hours, as commonly used to understand investment, represent a major construct validity problem. There’s nothing inherently wrong with measuring “time we spent on things” (hours) to understand better where effort is going, provided you make it clear that:

1. Time is not a fungible thing that can be infinitely allocated/re-allocated. This holds across skill sets, team context, and even across a “normal day”. For example, the “magic hour” of uninterrupted morning productivity can be vastly more productive than trying to wrap your head around something complex between 4 and 5 PM. The developer deep in the onboarding space, an expert in onboarding analytics and the user journey, working primarily with what customers see/interact with, can’t be immediately swapped into a deep, legacy backend refactoring effort.

2. Where you allocate time tells you nothing about the “quality”, efficiency, or efficacy of that expenditure of time. The highly skilled developer reactively called in to debug an area of the code they have no experience with will not have the same leverage as someone more junior, but more contextually situated. Ten people spending ten hours each is not equal to one person spending one hundred hours. Sometimes the entire system is constrained by a single specialist, a single decision, a single dependency, or a single review. Spending time elsewhere has negative leverage.

3. The context switching tax is real. Most efforts to understand the allocation of time, whether explicit or implicit, leave out large time investments. Imagine alternating between two tasks every ten minutes for a six-hour day. You’ll spend three hours (at least) on context switching and calibrating/orienting around the new task. Coordination is multiplicative, while time accounting is additive.

4. The relationship between time and outcomes is non-linear. 2x-ing where you spend time doesn’t 2x the outcomes. You might bang your head against a problem for eight hours and get nowhere, and the ninth hour “unlocks” all the value. The metric treats the input as continuous even though the production function contains thresholds and fixed setup costs.

5. Where we invest focus can have a long-term effect. Imagine spending 5% longer on every enhancement to pay off a bit of debt and “garden” the codebase. Three years later, a version of the team—some people have stayed, but some have left—is absolutely flying. Where do you “book” that time? And how do you connect that investment to the long-evolving, game-changing “return”? Some hours multiply future options. Some hours limit future options.

6. Time allocation schemes often force allocations into individual buckets, but there are typically causal relationships between those buckets. They aren’t mutually exclusive. One impacts the other.

Lean/Flow Metrics (and “story points”)

Helpful, when focused.

Lean/flow metrics include cycle time, lead time, practical WIP limits, and throughput (e.g., N stories per week). Most of these metrics have a storied history in manufacturing, which is both a positive—there’s actual math, theory, and practice behind them—and a negative when misapplied to software development.

Note the following:

1. In manufacturing, it matters what you are producing. Things that have sufficiently different work processes, value profiles, etc. are always divided out. This is sometimes referred to as “classes of service,” and it is a critical component of using these metrics effectively. Without that segmentation, changes in throughput, cost, or efficiency may simply reflect changes in the mix of what was produced rather than actual changes in performance.

2. A lot of product work is more akin to R&D, experimentation, and initial “design” than to manufacturing and mass production. In this setting, “waste” isn’t an inherently bad thing. If you try five things and one works, then you’ve purchased the knowledge for that fifth thing.

3. What constitutes a work unit is also mushy. Say you have a more open-ended stream of experiments meant to move a metric; what do you measure? Experiments per week? The lead time of the whole stream? Both? Where do you put the opportunities that get discovery work, but ultimately don’t get greenlit?

4. When we talk about “capacity” in manufacturing, you are talking about the sustainable ability to meet a particular type (or types) of demand. “This factory can produce N of this variety of shoes every month.” You invest in a factory that can meet this forecasted demand, with optionality to scale up the factory based on new demand. So, going back to some of the issues with hours, capacity is not (always) fungible and is not something that is “spent” at the time of use. It is an emergent quality of the factory’s design.

5. The boundaries of the system matter enormously. “Lead time” from when? Customer request? Commitment? Ticket creation? First code change? And when is something done: merged, deployed, adopted, or producing value? Moving the boundary can dramatically change the metric without changing the work.

6. You have to be ready for the uncomfortable reality that “work” spends a majority of its time not being “worked on”, as well as facing all of your bad habits when it comes to keeping people busy.

7. Many of these metrics describe a relatively stable flow system that remains stable over a sufficiently long observation period. If the system is rapidly changing and has a lot of variability in work types, you can be lulled into a sense of apparent precision.

8. Don’t get me started on story points. Any effort to use story points for anything beyond a disposable tool to spark discussion, right-size, or reality-check a near-term cycle commitment is a dereliction of duty. We have decades of evidence to support this.

The main thing with manufacturing is that it is essentially a demand-based pull system. The things being produced have a price, and there is a clear distinction between “sitting on the lot” and “in the customer’s hands.” That makes the economics an order of magnitude easier to reason about. Theoretically, we could imagine features going unused as inventory, unfinished work sitting in queues as WIP, or capabilities built ahead of actual demand as overproduction.

But the manufacturing use case makes much more explicit assumptions about the “value exchange.” There is demand for a particular thing. The thing has an economic value. And value is realized when it reaches the customer.

“But how about throughput accounting!?”

Say you have a very stable work stream, with a lot of historical throughput data to draw from, with a very tight line to value, with relatively homogeneous units of work and stable demand characteristics; you might be close to saying something like, “We allocated N% of our capacity for that demand this quarter” using these metrics. This would more closely resemble throughput accounting.​

The caveat is that those conditions are not necessarily ideal in product work for various reasons: 1) you want to do work that might have longer-term leverage, 2) you might want variety, 3) you might want to shift things to meet emerging demand, and 4) some of the most valuable work may deliberately change the system rather than simply flow through it. The more dynamic, exploratory, and leverage-seeking the work becomes, the less comfortable you should be treating historical throughput as a stable measure of allocatable capacity.

All to say that certain shapes of work are a better/worse fit, and you have to be aware of the differences.

To be clear, I think a lot of these Kanban-style/manufacturing metrics, along with PR-level metrics like cycle time, review latency, and merge frequency, can be very useful for understanding the health of the system, especially when combined with classes of service/work. I consider all of these metrics/ideas “required learning” for product developers. But their primary use is for continuous improvement: helping a team optimize around a particular type of demand, understand where waste is accumulating, identify bottlenecks, and improve flow.

Everyone should read The Principles of Product Development Flow: Second Generation Lean Product Development by Donald G. Reinertsen.

Alternative

We will get into tokens and AI in a sec, but first an alternative to points and flow. The only way I’ve seen the ROI discussion work reasonably well is to keep the cost side deliberately broad and apply much more rigor to the causal model linking that investment to value.

1. Treat the team as the cost unit. Start with something like, “This team costs roughly $1.5M per year.” Don’t pretend you can meaningfully decompose that investment into hours, story points, or individual features.

2. Connect the team to a causal model. Each team should be linked to at least one model that connects actionable inputs and leading indicators to mid- and long-term sources of differentiated growth.

3. Clarify the demand the team is expected to serve. Be explicit about the types, sources, and variability of demand. This gives you some basis for judging whether the team’s capacity and demand are reasonably matched. If they aren’t, that becomes a separate question, or at least something you should call out explicitly.

4. Fund durable lanes of focus when possible. Invest in enduring areas of responsibility or opportunity rather than constantly funding one-off projects, features, or stories.

5. Maintain a business case for the team. Document the causal assumptions, validity of the proxies, expected returns, major uncertainties, and what would cause you to change the investment. The team business case should transcend projects. You know you’re doing it wrong when the whole rationale for a team was a project finished years ago.

6. Be explicit about the business/product lifecycle. An emerging idea, a scaling product, and a margin optimizer/commodity should not be evaluated using the same expectations or metrics.

7. Judge return through movement in validated proxies. Look for evidence that the things you believe lead to differentiated economic value are actually moving, rather than trying to attribute revenue to individual units of work. Always ask yourself, “Are we more or less confident about this value hypothesis?”

8. Use operational metrics for their intended purpose. Use flow metrics for continuous improvement. Use hours to understand time leakage and broad patterns of effort. But don’t use either as the proxy for investment itself.

I still can’t get over the fact that companies rationalize $1m+ “teams”, but then choose to bury their value under layers and layers of pyramids, programs, points, etc.

For more on this approach, some topics of interest might be causal inference, econometrics, measurement theory, structural causal models, proxy validity, Bayesian updating, leading indicators, and real-options thinking.

Tokens

Imagine for a moment that someone in finance greenlights investing $50,000,000 in AI spending (measured as input and output tokens). Let’s first compare that to an equal investment in “human capital”. Consider all the gotchas and open questions:

1. Organizational capital / positive externalities. Employees don’t only produce assigned outputs. They also create knowledge, relationships, routines, and institutional memory that benefit other work. This matters much more when AI is augmenting people than when it is simply replacing a task.

2. Fixed vs. variable cost structure. Salaried labor behaves more like a committed/fixed cost over the relevant planning horizon; token consumption behaves much more like a variable cost. That distinction becomes especially important as inference gets cheaper and more commodity-like.

3. Capital formation / residual value. What asset remains after the expenditure? Human work can increase human and organizational capital. AI consumption may leave useful code, data, workflows, or intellectual property, but only if the organization actually captures it. Otherwise, much of the spend is simply consumed.

4. Marginal versus average returns. The first $1M in AI spend may be extraordinarily productive because it targets obvious, high-value use cases. The fiftieth million may be going toward increasingly marginal applications. This gets even trickier if higher-priced models are being purchased for differentiated capability.

5. Neither human capital nor AI spend produces value in isolation. Their returns depend heavily on the complementary assets around them. In augmentation scenarios especially, the relevant system is often human + AI, not AI by itself.

6. Both suffer from the long-term costs of overproduction. Producing more than the system can absorb creates inventory, maintenance, coordination, review, and cognitive overhead. Cheap generation can simply move the bottleneck downstream.

7. A cheaper/faster build is not necessarily a cheaper asset. In software, most economic consequences occur after the initial act of production. A compelling substitution case at build time may still have poor lifetime economics.

Are they comparable?

A lot of these points hint at the same two tensions. Will AI ultimately augment or replace humans? And will it provide a commodity input that people add value to, or will it be a premium capability? Or both?

What is the dominant mental model at your company? How did it get that way?

We have an unusually precise measure of one input sitting inside an unusually uncertain economic system (tokens). We can know exactly how many tokens were consumed and what they cost. Tokens give you a beautiful meter and invoice, which can create a much stronger illusion that you have a denominator suitable for ROI.

But we are still reasoning through much fuzzier questions: Augment or Replace? Commodity or Differentiated Capability? Or are we paying for differentiated capability? What complementary assets does it require? Does it create durable organizational capability? Does it increase maintenance costs? Move bottlenecks? Encourage overproduction? Change the economics of the work over its lifetime?

What theory of the system are we betting on

A serious AI business case is less “What’s the ROI?” and more “What theory of the system are we betting on?”

1. We believe using AI for __________ will reduce __________ without materially increasing __________.

2. We believe AI will primarily augment/replace __________, resulting in __________.

3. We believe spending __________ on AI will be economically preferable to __________ because __________.

4. We believe making __________ cheaper/faster will not simply move the bottleneck to __________.

5. We believe increased AI-enabled production of __________ can be absorbed without creating excessive __________.

6. We believe the primary long-term asset created by this AI investment will be __________, rather than merely cheaper production.

7. We believe the human + AI system will outperform __________ on __________, after accounting for __________.

8. We believe the downstream costs of AI-generated __________ will remain below __________ over its useful lifetime.

9. We believe paying a premium for __________ will produce enough additional __________ to justify the premium.

10. We believe this investment works economically only if __________ remains true and __________ does not become the limiting constraint.

So rather than:

We will spend $10M on AI and generate $30M of value.

the intellectually defensible version is closer to:

We believe this AI spend will change this production system in these specific ways, under these assumptions, with these expected benefits, costs, externalities, and downstream effects.

Then ROI is almost the output of the hypothesis, not the hypothesis itself. This is extremely important given that everyone is learning at the moment. No one has figured this out.

Conclusion

I’m struck by one thing. Imagine for a moment that we understood team funding as a combination of salaries and AI spend. A team might look at itself and say, “Geez, we cost $1.5M a year in salaries plus $1M in tokens. What is our business case here?”

I understand that this is somewhat utopian. In many companies, the goal is explicitly to use AI to eliminate some portion of that team, its managers, or the surrounding organization, based on some substitution math.

But if you think about the problem at the level of the team + AI system, the solution is remarkably similar to the “alternative” described above. What demand are we serving? What causal model connects our work to differentiated value? What durable areas are we investing in? What assumptions have to hold? What evidence would make us increase, decrease, or redirect the investment?

In fact, this kind of business case may be a prerequisite for giving teams meaningful responsibility for AI spend. If you want teams to make intelligent tradeoffs between people, models, tools, and other forms of capacity, they need a model of the investment they are responsible for and the value they are trying to create.

The tensions point the way to the real decisions. Most of what we call prioritization is forging agreeable narratives. It resolves very little and produces very little real commitment. But if you pay attention to where the same arguments/tensions keep surfacing, where conviction keeps failing, where momentum keeps winning despite everyone’s better judgment...that’s where you actually need to make the hard decisions.

There’s something Gestalt about this: you already perceive the whole shape of the problem. Your intuition is already doing that work. The framework just forces you to decompose it back into parts and lose the signal.

Later on in this post I share a simple, actionable exercise you can do with your team to surface these tensions. But before that, let me lay out my case.

If you like the newsletter, consider…

Leave a Coffee Tip!

No One Retros The Framework. Why?

There are two parts to prioritization:

1. Prioritization judgement. What deserves attention and investment?

And then:

2. Prioritization-as-enacted

What would actually cause things to move in alignment with those priorities.

A lot of teams use prioritization to AVOID what is sitting right in front of their face. Instead of having the harder discussions, they slip into perpetuating the same old fake rigor and safe decisions.

I rarely, rarely talk to teams and hear “Oh well, during that prioritization exercise three quarters ago we really dropped the ball on prioritizing that because it was 3 not a 4.” Or “Woah boy, we really should have allocated 60% of our time there instead of 41.23% based on the CoD calcuation.”

I almost always hear something like “We just didn’t have the conviction to stop _______.” Or “We just kept plodding away on [something decided on two quarters before that] and everyone knew that was wrong.” Or “You had 20 teams saying they were slowed down by X, and we still didn’t have the political will to prioritize that measly platform investment.” Or “Let’s admit we were wrong about [that thing] and move on.”

The conversation in the retro is never about the framework. And always about how people rationalized, communicated, showed up, and made sense of things together.

Teams will spend hours/days/weeks trying to invent some sort of “defensible” system, when massive asymmetries and “problems to prioritize fixing” are sitting beneath their noses.

You don’t need more options! Or more spreadsheets!

The example I use is how as individuals we’re often swimming in the things we should be doing, the self-care we should be performing, the nagging tasks holding us back, and the big threats looming on the horizon that we can’t quite mobilize or motivate to tackle. Sitting down with a spreadsheet and trying to “reason things out” doesn’t solve the problem. The same with teams.

Most prioritization discussions obsess over prioritization judgement. But it is the recurring argument around prioritization-as-enacted that can be incredibly informative.

Instead of:

“Let’s prioritize again”

You instead should ask:

“Why has our existing prioritization process repeatedly produced this particular tension?”

Theoretical vs. Enacted Prioritization

The missing thing isn’t belief. It is commitment under present-day inconvenience. Good prioritization isn’t simply choosing the right work. It is recognizing the traps and tensions that prevent your portfolio of bets from panning out the way everyone claims it should.

I can’t count how many times I’ve been in a room where there seems to be a lot of agreement about some far-off threat—something we should do, something existential even, but not immediately existential. But for various reasons, despite the “priority”, the team can’t mobilize any sort of conviction. Maybe it will involve too many teams. Maybe there is currently “no clear owner”. Maybe there are differing opinions on how much energy to commit based on perceived levels of problem and solution certainty.

A lot of time, the key problem is that the problem is intangible. It is a future threat that people find all sorts of ways to rationalize will somehow work itself out. Every pre-mortem singles out the threat, and every quarter there are half-hearted efforts to take it seriously, but nothing happens.

When their guard is down, I’ll ask about some sort of existential threat looming on the horizon. There’s rarely disagreement about what that thing is. It doesn’t need to be rationalized. If there is disagreement, it is between markedly different futures with existential issues, or a comparison of existential problems (we need to live to even worry about that). Or magnitudes of disruption. The problem is imagining what action right now might look like. It isn’t some sort of rational, measured, economics-informed debate.

Is this a prioritization problem?

Well, yes, in many ways it is. But it isn’t the kind of rational pros vs. cons, risk, upside, effort conversation that people normally associate with prioritization.

Similarly, I can’t think of any work situation I’ve been in where there isn’t a priority people love to work on. The value is tangible. No one questions it. It is fun. It feels meaningful, at least on the surface. Saying yes has become a well-worn, habitual knee-jerk reaction, and no one is really complaining. In a meeting, trying to defend your value? Mention this priority. But there’s also a widespread intuition that it might be too much and that it is getting in the way of something far less comfortable, and maybe way less easy to rationalize. There aren’t enough guardrails. You’re running a very serious risk of exploring the wrong side of the effort/outcomes curve. No one wants to be the person to say “we can have a bit too much of a good thing”, but you really need someone to say that.

Is this a prioritization problem?

Again, yes. But because the work is happening right now, and because teams have a tendency of avoiding discussing the priority of “work in flight” (and because people get used to fun things that feel even a bit meaningful), you’re starting to drift from a standard prioritization motion.

Tension Patterns

Those are two motions. Here are some others:

• Areas where you need to pass along a “default yes” because prioritizing all the individual items will not result in anything getting prioritized (the meetings are more expensive than just cranking)

• Areas with a ton of inertia where someone needs to call a big reset to save the effort. The reset IS the item to be prioritized

• Efforts that have been almost-done forever and just need one concentrated push to finish (different from the reset because you know the direction is right, you just need to stop tip-toeing around the close)

• Priorities that keep boomeranging because there is insufficient data/discovery, so people keep saying “when we have more data”

• Early-stage efforts that need care and room to grow, but not a free pass. The big decision is to really commit to planting the seed, and caring for it (without falling into “innovation lab” or “hire to get anything done” syndromes)

• Effort blocked because you don’t have the “capability” yet. The big decision is whether to (and how to) build those chops

• Important efforts where there’s a credible path to radically less scope that still captures most of the value (this is more of a lens, it can apply to almost any of the above)

• Priorities that feel urgent because “the window is closing.” But is the window real?

• Modest-looking efforts whose real value is indirect: they make a dozen other things faster or better. On paper they should win every conversation, but they get bogged down because the priority is less sexy, and everyone wants the benefits without funding/helping that shared capability (crisis of the commons)

There’s a polarity between many of these. You can’t do everything at once. You can’t cap the comfortable thing AND do the consolidated push AND plant the seedling AND fund the capability gap all in the same quarter. You keep a tight portfolio of risk profiles across time, value, and urgency. But it is not nearly as mathematical or rational as we think. It is more like managing tensions: knowing which ones to hold, which ones to release, and which ones you’ve been avoiding.

My theory is that even when we try, we’re not doing any sort of “rational” prioritization. But instead of fighting that, you can actually use prioritization tensions as a signal on the way to a more balanced portfolio of bets.

Conventional discussions of prioritization assume something like: “We have several plausible things we could do. Let’s compare value, risk, effort, confidence, etc. and decide which ones to invest in.”

But in my experience, the real problem starts AFTER that. We often know what we should do — that isn’t the problem. We know what the existential threats are. We already know we’re overinvesting in some things. We already sense which projects need a reset. What we’re struggling with, or don’t have the conviction to do, is how to create the organizational conditions required to act on that knowledge.

Meanwhile, you’re currently doing a lot of things you should either not be doing, should spend more time on, or should radically change your approach. The challenge is two-part: yes, you need to be considering the right or less wrong options, but it is also HOW you approach those things.

Exercise

Try this with your team. Fill in the blanks. If you can’t fill one in, skip it.

1. “We keep coming back to _______, but no one has a clear solution. And we seem unwilling to jump headlong into it. Every pre-mortem singles it out. Every quarter there are half-hearted efforts. But the threat is just abstract enough that there’s always a plausible story for why it might _______. We keep pushing it off to _______ and hoping that’s enough.”

2. “No one debates that _______ is important, and that is part of the problem. It’s fun. It feels productive. Saying yes is a well-worn habit. But unless we set limits, we’ll keep polishing it forever, well past the point of diminishing returns, while _______ starves. We’ve probably already captured _______% of the value.”

3. “Unless we agree to do _______ in principle, we’ll keep wasting time trying to prioritize each and every little piece of it. No single item is big enough to survive a prioritization meeting. But collectively? People should just be able to pull from this pile whenever _______. The value comes from doing it in _______, not batching it into some giant future initiative.”

4. “Everyone in the room knows something is off with _______. Confidence has dropped since _______. But the momentum is crazy. Ask a question, you get shot down. Nobody wants to be the person who held up the train. What we actually need to prioritize is permission to _______ it.”

5. “_______ has been almost done since _______. Everyone is tired of it. We keep giving it scraps of attention, and it keeps lingering. _______ stays blocked waiting on it. We need a short-lived period of extreme focus, or we need to stop pretending. That means explicitly saying _______ slows down for a bit.”

6. “We keep discussing _______ but never actually learn anything new. Someone is always ‘doing discovery’ on it amidst _______ other things, but nothing real emerges. We need to spend real, concentrated capacity resolving _______ before we can move.”

7. “_______ needs room to grow. We can’t judge it by the same standards as _______. We need to protect it and give the team leeway, but also hold them accountable for _______. The question isn’t ‘Is this paying off yet?’ It’s ‘Is this developing in a healthy direction?’”

8. “We keep calling _______ important, but no one will fund it, stop _______ for it, or hire against it. Is it important enough to displace _______? Important enough to add capacity for? Or neither? We either invest or say no.”

9. “What’s really stopping us on _______ isn’t the work itself. It’s the investment to be able to do it. We keep wishing we had _______, but we haven’t prioritized building that muscle. We have to prioritize our ability to do this, not just doing it.”

10. “The scope on _______ is huge. But there’s a credible path to radically less scope that still captures most of the value. What if we just did _______? There’s an 80/20 here if we’re willing to seriously test _______ before committing to the full thing.”

11. “_______ feels urgent because _______. But is that window actually real? What specifically changes if we wait until _______? Sometimes the answer validates the urgency. Sometimes it exposes something people just want to do now.”

12. “On paper, _______ looks like just another effort. But if you consider everything that would benefit, it’s critical. _______ gets faster. _______ gets cheaper. Unless you explicitly account for what it multiplies, it will always lose the prioritization conversation.”

Each one of these mad libs is designed to surface a tension, and a potential prioritization decision. Use them before you use whatever framework you had in mind to get the real questions and decisions on the table.

These are the real tensions to resolve.

I wanted to send a huge thank you to my paid subscribers. They don’t get anything in return but still support the newsletter. Until recently I didn’t have a way to do a one-off tip. But now I do. I you like the newsletter, consider…

Leave a Coffee Tip!

North Star Framework

“Should I try to implement the North Star Framework with my team?”

(Note, I wrote the North Star Playbook while at Amplitude, with Jason Scherschligt)

North Star Framework is an interesting idea, but it is probably most valuable as a teaching tool, not as a practical practice. That isn’t to say that some companies don’t “do it”, because they do, but that it is more valuable when it comes to teaching things like leading/lagging metrics, actionable inputs, building (and testing) causal models, avoiding vanity metrics, and the “art” of crystallizing a unique strategy for your company vs. playing a standard game.

When I check back with companies/individuals who attended workshops, it is very common to hear something like, “I learned a lot, and we sort of adapted many of the ideas. We have multiple KPIs, leading and lagging, and some things aren’t as worked out.” The people who didn’t get value out of it tried to install it as a framework and eventually hit resistance with the inevitable edge cases, exception handling, etc.

This is all to say that what follows will work in some of these ideas, but will not be about the North Star Framework.

(Here is a Miro activity library for NSF)

Trees/Pyramids

“Should we build out one of those vision-to-tactics pyramids to get everyone aligned?”

Next, this might sound heretical, but I don’t recommend “pyramids” that build a tree starting with vision, strategy, tactics, etc. I know they are super common, but I find that they often end up distracting people. If someone owns something, and if people are working on it / working towards it, then by all means make it clear. But most companies end up with a bunch of buckets and maps that seem wholly designed to “simplify” things for leaders, and to abstract the idea that lots of real work is happening.

There’s nothing wrong with label taxonomies, but people start to imagine that those labels are real things. Most of the time they aren’t—they are roll-up mechanisms, and seem mostly designed for 1) deckware, 2) and a false sense of organizational symmetry or “clear” roles and responsibilities. Sure, make a pyramid slide if you need that kind of thing to leave leaders feeling aligned, but then put it to the side to start generating impact.

Put another way, there’s only one place that “the work” is happening, and that is at the front-line team level. It may have gotten there through layers of the org, but the work—the problem solving, creativity, hands on keyboards, interactions with customers (hopefully), etc.—is happening on the front line. Real strategy is messy, non-hierarchical, might span days or years, and doesn’t fit into neat boxes, so don’t waste time trying to do that (or at least don’t waste too much time).

Stable Teams?

“Everyone says we need stable, durable product teams. Is there a downside?”

There has been a lot of talk about “stable” product teams, but I wanted to dig into the dark side of this. A lot of companies are struggling because they build internal kingdoms around “stable” teams. They rationalized these highly durable capabilities that would exist until eternity, based on sort of irrefutable logic (we’ll “always need that”). Or, they wanted to build something, a short-term project, and everyone was busy, so when money was plentiful they just spun up a “new team” to build it. And then a director. Then a GM. Before you know it you have three layers of hierarchy built around “products” that didn’t really exist.

And then…the only way to unwind this is layoffs that really impact people’s lives. Not because you couldn’t, in theory, re-assign those folks to different areas—because you could—but because the physics of the org chart, and how layoffs work, necessitate you to blindly unravel your prior org chart decisions.

Which is all to say that “stable” teams and “durable” ideas are only useful until they aren’t useful, and sometimes knowing the difference is hard. Of course, leaders will not say this outright (or at least the CEO’s who speak publicly about layoffs). They might make vague mention of bureaucracy, the “work around the work”, etc. But the reality is they scaled too quickly when money was more plentiful, and now need to rationalize how to unwind all of that.

Shipping Smaller/Faster

“How do I get my team to ship smaller and faster?”

In my experience, it is way easier to teach a team that can work small to think big, than a team that thinks big, to work small. Both are possible, but they require different approaches. A team that can deliver continuously, regularly, get frequent feedback, and generally keep the tempo of changes, assumptions tested, and risk reduced HIGH, is a godsend. It is a gift. If you have fostered an environment AND the skill/experience landscape to make this possible, you should always cherish it.

Now, that sometimes means the team will be short-sighted, and may iterate to nowhere. You run the risk of having a “functional” feature factory with broadly usable things that don’t really move the needle, but if you can do this you’re well on your way. As mentioned, this is part environmental. You can absolutely have people who are good at this at their last company, completely flounder if you’re constantly interrupting them, changing the strategy, loading them up with WIP, and always asking them to prematurely converge on estimates, etc. Environment matters.

But creating that environment AND working small are skills you have to hire for, and nurture internally. Put another way: you can’t make shots you don’t take. The team has to be able to get changes out into the world and HAS to be able to measure impact. They also need chops at rapidly considering options, and figuring out how to slice up those options to get the most leverage possible.

Cycles

“What’s the right planning cadence? Quarters? Sprints? Shape Up cycles?”

Quarters are too long to help you work small, and too short for the bigger, meaningful things you want to do. 2w sprints are too long to help you work small, but too short for the big meaningful things you want to do. 6w is actually a decent forcing function, because you can fit some pretty darn meaningful things into 6w, but if you are operating a company at any sort of scale, you might feel a bit of whiplash thinking only in 6w bets. It is doable, but you’ll find yourself 1) tackling things that are actually bigger than 6w on occasion, 2) losing a sense of the more durable “lanes” in the 1-3 quarter range, or longer.

This is all to say that time is WEIRD. It always feels too long, and too short. There are always opportunities to work smaller, and there are always situations where working small misses the “bigger picture”. But here’s a bit of philosophy: life is only happening NOW. We are navigating a landscape shaped by past decisions, and we are forging ahead into a partially knowable future, but at the end of the day what really matters is what is happening right now.

Why is any of this important? First, don’t get beholden to any one cycle as “the” cycle. You are almost certainly missing something. Second, don’t let your selection of cycle get in the way of working small (or thinking big). The number of teams I see chasing their tails because of some arbitrary quarter cycle AND magically making all their bets take one quarter is insane. SHIP! LEARN! Don’t wait around for a quarter.

If quarters are causing some sort of wholesale rebalancing of effort, that is a sign you 1) probably need to be working smaller AND 2) probably have some tough prioritization questions to make in the longer timeframe. Instead, people treat these cycles at any duration as “boxes to fill”, and it is incredibly damaging. A quarter is not a big sprint. It never was. And any effort to scale up “sprints” to “quarters” to make them somehow more corporate-compatible is a perversion of the idea.

Lanes

“What does it actually look like when a team is really humming?”

All the best and most productive/effective times in my career have followed a pattern. A handful of somewhat stable “lanes” with a couple metrics and directional sense of the opportunity, and then a team just cranking it out week after week. What happened last week? What’s happening this week? What’s working? What isn’t? You’ve probably felt the same thing at some point.

The puzzle here is that despite how salient this all feels, it can be devilishly hard to recreate these conditions. People look at all sorts of frameworks, devices, tools, tricks, etc. yet the actual “feeling” is way simpler. It feels like: a tension between the more stable, and the more rapidly changing. It feels like: a strong bias for action and movement (even if that action is discovery/learning). It feels like: talking about things you DID and what you learned, and less about talking about all the things you plan to DO, and all the things you hope to LEARN. 80% of the energy is on the here and now.

This doesn’t mean that everything is short-term and myopic or “purely tactical”, because you’re like a mountaineering group navigating a mountain where you are always scanning at multiple resolutions AND choosing your next step wisely.

Why is any of this important? Teams will go to the ends of the earth to avoid this motion because it is scary. They’ll invent all kinds of work to get in the way of the sheer, raw, unfiltered act of getting stuff done, and holding each other accountable. And course correcting based on what they’ve learned. They’ll spend weeks and months “going back to the drawing board and building alignment” because they are scared about the sheer act of doing, and potentially marching off in the wrong direction.

So it is like this elusive “vibe” that you know when it is happening, but can be hard to get into for various reasons—with all sorts of excuses we make to ourselves and to each other about why it isn’t possible. I don’t think this is laziness. It is very human fear. It is the fear of forward momentum under conditions of uncertainty. It is the fear of “picking the wrong things”. It is the fear of showing up at the meeting and saying “oh yeah, a million things came up last week, we need to have the uncomfortable discussion about whether I should have said NO to those things.”

Humans doing normal human things in the context of the team creates these conditions. So the burden is that we’re ALWAYS trying to converge on this vibe, and always slipping out of it. That is the work—of ICs, non-ICs, everyone.

Deciding On How You Work

“I’ve got a bunch of senior people who can’t seem to agree on how to work together. What do I do?”

Experienced teams/people are especially prone to this: you can get a bunch of smart people together and half the time is spent just reconciling different variations of how they like to work, and their quirks. People can be terribly “nice” about this and try to fill in the gaps. But experienced people, especially, often just want SOMEONE to decide how things will work. It isn’t because they can’t figure it out—they can—but all the waffling and gap filling, and back channel influence can be a huge drain. Can someone PLEASE just define the game and the playbook and let’s get moving!

This doesn’t mean everyone needs ultimate certainty or needs all the t’s and i’s dotted (experienced people rarely need that), and with water under the bridge you rarely need things concrete and explicit, but when you get a group of smart people together who haven’t worked together, you should bootstrap the whole thing with “V1” of a way of working that people generally agree with, and get moving. Leave all the theorizing for the future. The team can adapt as it goes along.

Note: I’m not talking a dictatorial motion here. You could collectively decide to get in a meeting, write some initial working agreements, and go. Or you could say “Hey Bob, I hate to do this but can you just hammer out the V1 for how we’ll work”. Either way is fine. My general point is that teams waste so much time and brainpower trying to arrive at a reasonable game, and I think it is better to just start with V1 and go.

The big risk here, of course, is that the person trying to figure out this game is trying to oversimplify it, deck-ify it, or make it palatable for the people who aren’t actually doing work, which ends up meaning you have a starter OS that doesn’t really reflect how people work. It is a showpiece. You don’t want that.

Estimation

“How precise do our estimates need to be?”

You don’t really need to estimate at resolutions more than 1-3d, 1-3w, 1-3mo, 1-3q, 1-3y. That’s it. Anything more is a waste of time. Importantly, be very clear about the “thing” and make sure you’re working at the atomic level. You can have a big messy project that is 1-3q, that while being amenable to “breaking down” is just what it is. It is a 1-3q bet.

It might help you to imagine a bunch of 1-3w and 1-3mo bets that “add up” to this big thing, but the important thing is you have a 1-3q bet, and you’ll have to figure out how to work small, de-risk often, and use your company’s money wisely. You could also have a 1-3q bet that is more open-ended, amenable to experimentation, has a bunch of pivot/proceed points, and can be funded iteratively based on outcomes and progress. OK. In that case it may help to think of it as “housing” a bunch of 1-3w, 1-3mo bets. That makes more sense.

But it is an inherently different relationship between the “things”. You don’t need to finish all the “things” for the big thing to work, whereas with the 1-3q big messy project you probably will need to make significant progress across the things. In the experimentation friendly example, you might only need the first couple. In the big project example, you don’t need all the details, but you do need a strategy for de-risking the thing over time.

With both, you can STILL meet up every week and talk things through, and you will still (hopefully) have inertia every day towards the goal, but it is a different motion. So: don’t tear yourself up with estimates.

Lanes Revisited

“How do we organize ongoing work that isn’t just a list of projects?”

I had a mentor once, a product leader, who just showed up one day and said “Look it is all about the lanes.” He was referring, roughly, to more durable, outcome oriented swimlanes. They might show up on a roadmap, on a slide, etc. This was a deceptively powerful statement when you understood how this played out in practice. We’re not talking kanban lanes (which tend to depict a workflow), and we’re not talking about value streams, though the idea is similar. Sure, you can use a lane for a long project, but that isn’t the basic idea.

But basically: “Hey, for the next couple months or quarters we’re going to keep these couple things stable, revisit them, measure them, and let them guide the way. And when they stop making sense, we’ll change them or pivot.” This seems so basic, but it isn’t. It is far easier to come up with hand wavy strategic pillars, or generic “buckets” that all work falls into. Or to come up with discrete little projects. It is much tougher to find that perfect balance of durability, outcome-oriented, actionability, non-fluff, customer impact, causal model, etc. that a team can ACTUALLY work with.

This was one of my learnings with the North Star Framework. One of the real benefits of the NSF was the idea of the actionable inputs and really tweaking those until you had some that held together and withstood the test of time (but weren’t so open ended, or so “bucket-ish”, that they didn’t really guide any action). The idea is somewhat fractal, although any time you start making trees of things, you have the potential to water things down.

Imagine a company with 150 teams. That is about 450-600 lanes, roughly. With some lanes for the big stuff that has to cross teams. “Oh god,” the leader says, “That is so much. There’s no way I can possibly go into a meeting and discuss 600 things.” To which I would respond this way: your lanes are costing you millions of dollars, or at least in the high six figures (think a team might have a couple that stay stable for a year). Are you really telling me that you can’t bother with understanding your million dollar investments? “Oh, but we will when we roll those up,” … to which I respond, “Oh really, to the extent that you’ll pressure test these very real bets?”

Here’s my take. Just admit you have 450-600 durable lanes, and use AI, routines, and rituals to hold yourself accountable. Use label taxonomies if you really need to roll stuff up. But realize that the work is happening at the lane level.

Shapes Of Work

“Can we treat all our initiatives the same way and use one template?”

Lean into the fact that things have different shapes: different risk profiles, different levels of experimentation friendliness, different “sizes”, etc. I’ve never done a shape exercise with a team and ended up with anything lower than 5-8 key shapes of work. This is not about being overly pedantic and categorizing everything, it is more about never imagining you can see initiatives, epics, bets, or whatever as apples to apples, and resisting efforts to standardize or oversimplify.

I’ll give you an example. Most companies have a launch tier where product marketing describes the nature of the launch from a marketing perspective. But guess what? Some things are going to be incredibly disruptive internally, and require a ton of coordination, but marketing isn’t going to make a peep out of it to customers. In fact it may be critical for customers not to know anything about the launch. The “simple” distinction between internal blast radius and external launch demands has confused so many companies I’ve worked at, or advised….just that tiny distinction. 9/10 someone in marketing would say “yeah, but we need to keep it simple”, and their categorization scheme would pervade with executives and others.

Don’t do that. It doesn’t mean that everything needs to be analyzed to death—but know that the work we do defies simple categorization schemes and labels. To get around this: don’t use templates. Look at the THING. Read it. Write about it. Be explicit. Let people raise the nuances. Surface the differences that really matter. And do anything you can to not flatten everything you do down to someone’s slide deck. Sure… you may eventually need the slide deck, but that isn’t reality.

Routines (Creatures Of Habit)

“How much process does a team actually need?”

Routines: it doesn’t matter if they are explicit or implicit, but have them. Great products are built on routines. They can be very fluid. They can be more strict. But they are there. These collective habits are what makes great work happen.

The key problem is the degree to which routines either 1) get processified to the nth degree, all but eliminating any chance that they are effective, or 2) are so implicit and so loosey goosey that when something changes, it makes it hard for the team to reset quickly. A group of experienced people can go very far with just “meeting when it makes sense”, but even experienced people are vulnerable to the traps. They are vulnerable to the traps of not raising risks because it is uncomfortable. They are vulnerable to the traps of avoiding hard conversations, of getting carried away with status quo bias and not challenging their direction when it makes sense, and putting off the tough call “one more week”, when they really need a coworker to set them straight (or at least try to).

Routines are the heartbeat. And I think to some degree we are creatures of habit. We start enjoying our routines when they help us regularly. You can make the agenda variable, but it pays to get people in a room (virtual or otherwise) to connect.

Everyone has been on that team that starts with best intentions, has a great weekly meeting, stays on the gas, and then as elephants in the room accumulate and priorities fracture, they start to lose it. Leaders spend more time in other rooms planning for something far off, than connecting on what is happening right now. People do 1:1s to plot out what’s next after the impending collapse everyone knows will happen. All to say that if you can push through—push through that thing that happens four weeks in where either 1) it becomes a zombie meeting everyone sleepwalks through, or 2) it stops happening all together—then you’ll be better off.

AI

“How is AI going to change how we work?”

AI? AI will turbocharge your bad habits and turbocharge your good habits. My hunch is that 80% of the world stays exactly the same, all the principles apply, and now you have this tool to do better work in all the ways you know how (hopefully). You could use it to do things you know you should have been doing this whole time, but didn’t have the time in the day. Practices are changing every week. One week it is ____ engineering. One week it is ______ engineering. Back and forth. So yes, you will need to innovate on the practices, but expertise still matters / chops still matter.

Example: say you have an org that was running a feature factory before AI. BETTER feature factory. Say you have a team that was thoughtful about collecting research but was short-staffed, and couldn’t keep up with organizing their research library. Bingo! Better at all that. Amazing. Or you knew you needed to instrument analytics, but never got around to it. Write a skill. Instrument more analytics. Boom!

The real danger is that some people don’t know what good looks like, or have never worked in an environment that did truly high leverage work, and that’s setting the bar for them around what they think AI can help with. Another example: synthetic personas. Before AI I could create a synthetic persona. It took a bit, but it was doable. With AI I can make terribly persuasive synthetic personas. Is it a good idea? For what? For what boundaries? For right now, at least, only skill will give us an answer there. It is going to be a wild ride, and practices a year from now will likely look very different (while we’re potentially digging out from a huge crash and picking up the pieces).

Outcome-centricity (Misunderstood)

“Everyone says be outcome-oriented, but how do you actually do it?”

Outcome-centricity is hard, and I’m tired of people describing it like some kind of mindset, etc. There’s a reason why the cart optimization team at an ecommerce company gets more love/attention, than some team doing design systems. It is closer to money. There’s no free lunch. You will need to build a causal model that starts with near term things you can influence, linking up to the long-term things that will propel your company forward.

Even harder, it isn’t all math, or even mostly math, because you are SHAPING the future as well as measuring the changes you introduce, and how they impact near term things. If it was an easy math formula, everyone would be doing it. It isn’t! It takes conviction, leaps of faith, etc. As I always say, if everything was an A/B test and you knew the lifetime value of every customer, and if every pattern that happens now has happened a bunch of times in the past, then you are a phone company, and you are a commodity.

I mention this because all the talk about projects to products and outputs to outcomes, etc. etc. is way too theoretical. It doesn’t get to how uncertain it is, and the mix of math, conviction, storytelling, and narrative you need to make it all work. So it leaves teams feeling they’re never doing enough, and completely negating the routines and practices—the staring at input metrics, revising, discussing assumptions, checking, monitoring, switching things up—needed to make it all happen.

I worry I’ve been a part of this, though I try to make a big point that you can think of outcomes at various time scales and resolutions, and that it is all a “model” that we keep refining. Or at least I try to.

Better Questions

“Do we need to have all the answers before we start moving?”

Related, what really matters is that you’re always asking better questions, not necessarily that you have all the answers. You can improve your ability to answer questions over time. But some teams do activities where they are GREAT at asking powerful questions—things that really land—but then they seem to go back to ho-hum business as usual. It is like they take off their “good question” hat because they’ve had the “oh you’re the type who doesn’t get anything done and only asks good questions” line.

Not coincidentally, asking better questions is a function, in a sense, of getting answers. Answers let you sharpen new/better questions. This all goes back to routines and tempo. You’re knocking out work on your lanes, you’re asking good questions, you’re getting some decent answers, you’re double-checking your core assumptions, you’re progressively moving the needle….that is far better than having the perfect answer in every meeting.

Good Goals

“How should we be thinking about goal-setting?”

Goals. Set them. But don’t be ruled by them. Goals can be set for anything: learning goals, shipping goals, outcome goals, goals on near-term metrics, goals on lagging metrics, goals on practices, goals on continuous improvement. The big anti-pattern I see is that people just go through the goal motions, and they don’t mean anything. Or they just dream up everything they want to do, and goals are sort of the “success metric” afterthought. Mix up your goals—prescriptive, descriptive, near term, long term, leading, lagging— and lean into the HABIT of goal setting, not the performance of goal setting.

More…

“Anything else top of mind?”

A couple more to make it 20:

1. Lean into anchor artifacts: stuff you write and revisit, vs. always reinventing the latest doc format, the last initiative brief, etc. Revisit. Do they still make sense.

2. Real progress can be sensed daily and weekly. Even if it is “big stuff” or “hard stuff”. It is all about coaching the team in how to break things up without losing the big picture.

3. I get tired of the “leaders are the only people doing long horizon” idea. Anyone at any level can be thinking about multiple horizons. More experienced people can often reason about longer horizons a bit better (or know when reasoning over longer horizons doesn’t make sense), but ideally you have those experienced people on the front-line. If not. Why? There’s gotta be some power ICs who love showing people how to do the thing.

4. I always say this: but powerful ideas imperfectly measured are more important than perfect measures for not so powerful ideas.

Boring Product

“What does good product work actually feel like day to day?”

Boring product. It is in the doing. Every day. Every week. Moving things forward. Thinking across horizons, but not being caught in the past, or dwelling in the future. It is a collection of habits. You can tell a mile away when something isn’t working for a team. Just listen. Do you hear it? Do you hear the motion? The learning? The connecting the small to the big? Or is it always like a movie that never seems to find its footing. Trust your instinct with that.

Hopefully this was interesting. I know my writing doesn’t seem actionable sometimes, but I do my best to get advice out there. My Achilles heel, which is probably evident in this post, is that I know a ton depends, and a ton of this is about nuance and rolling with the changes. Let me know if any of this helps.

If what you are mapping is incoherent, don’t fall in love with the map (or your personal ability to navigate with it). Use it to refactor what you see.

Here’s why.

More Coherent Organizations

When strategy, structure, technology, and incentives are coherent, an organization can leave a great deal implicit. A team and its mission become reliable shorthand for its goals, ownership, work, funding, and intended path to impact. These are technically distinct structures. For example, work isn’t the same thing as goals, and teaming isn’t the same thing as a product taxonomy. But when they map cleanly to one another, there is little need to name every distinction, map every relationship, or maintain separate representations.

You can look at the teams or the goals and reliably fill in the blanks. There is tons of context, but there’s little need to spell it all out. In a coherent organization, what you learn from one map remains useful when you switch to another. In an incoherent organization, what you learn from one map may stop being useful the moment you switch to another. You have to reorient, translate, and fill in the blanks all over again.

Coherence lowers the energy required to make sense of what is going on—the “context.” For humans. And for humans using AI.

The best metaphor I can come up with is a city with several map layers.

The layers might show streets, utilities, public transportation, neighborhoods, zoning and land use, school districts, political jurisdictions, and so on. All of those things are different. But when the structures mostly line up, you don’t need to keep naming which map you are using. If you say “downtown,” people can infer transit access, building types, land use, population density, and patterns of activity. Imagine flipping through the layers in an online mapping tool: each layer shows different information, but the layers seem to line up. When you are out walking around, things make sense.

Back to organizations.

Less Coherent Organizations

Imagine something critical to the business that involves eight teams spanning eight different product hierarchies, all with competing goals; silos across product, technology, and design reporting lines; unclear ownership; theoretically distinct customer segments; different impact hierarchies; and different funding hypotheses.

Suddenly, you’ve lost your bearings.

A consultant friend described navigating an environment like this:

“I’d meet with a team and be completely blindsided by what they were getting into. I had to sit there and go through all of their work just to make sense of it. I found myself constantly surprised: ‘You’re doing what? Why? For who? Who made that call?’”

This is like a city where every map layer cuts across the others differently. The school district boundary runs straight through the middle of a neighborhood. The buses follow one set of corridors, while the trains and bike routes follow completely different ones. The place everyone calls “downtown” is split across two municipalities, three zoning districts, and several utility providers. The nearest fire station technically serves somewhere else. Parking permits, building codes, public services, political representation, and transit access all follow different boundaries.

This doesn’t mean some people don’t absolutely know their way around. They may feel perfectly at home, know which unofficial paths to take, and reliably get what they need done. Sometimes you can even pay people to help: expeditors, tour guides, brokers, fixers, or longtime locals who know which office to visit and whom to call. Individual neighborhoods can also feel perfectly coherent, at least until you need to cross a boundary to get something done. But the fact that insiders can navigate a place does not make the place coherent. It may simply mean that navigation has become a specialized skill.

And this is the big puzzle.

A company can become extremely sophisticated at documenting dependencies, translating between groups, creating cross-functional forums, hiring fixers, and building tools that help everyone navigate the incoherence. Those things may be necessary. But they can also lower the immediate pain enough that the underlying incoherence never gets addressed.

AI will not magically make context coherent. It can surface and compare the maps, and maybe even call out incoherence, but it cannot make conflicting structures, incentives, goals, and definitions line up. In fact, when it smooths over those contradictions, it may exacerbate the underlying tensions while producing the appearance of alignment.

The goal is not perfect coherence. Perfect coherence would likely require either stasis or an organization so tightly constrained that it could not adapt. The goal is to notice where the maps have stopped lining up, decide which gaps matter, and continually bring the most consequential ones back into alignment.

If what you are mapping is incoherent, don’t fall in love with the map. Use it to refactor what you see.

(I have a question at the end, if you get there)

Picture a company annual kickoff.

Leaders are sharing a strategy that is objectively not a strategy. Not just by a purist’s standards, but assume for a moment that the experienced people in the room all know it isn’t a strategy.

So we have:

Believers

The less experienced people in the room who are happy that someone seems to have a plan. They’re not skilled enough to know what a strategy is. If it sounds somewhat coherent to them, they’ll be excited.

Players

The experienced people in the room who know it isn’t a strategy, but also know the “game” behind the real strategy. They know that the annual kickoff is a performance. They know how to play the funding game. They know where the “real” conversations happen, and the spreadsheets where those conversations get codified. But they also know when to nod, when to smile, how to nail their slide, and generally keep the show going.

We can probably split these folks into the hard-core Dark Triad people, the “you just don’t understand how hard leadership is” crowd, the “some things we just can’t say out loud because it would demotivate them” contingent, and the “savvy operators who know how to get stuff done, know this doesn’t matter, and are pretty decent people” crew.

Oh, and the “if we focus, people will phone it in, so our strategy should be ambitious and we should try to do everything” crowd. But that’s a post for another day.

Checked Out

Next, you have the people who have checked out of these types of meetings. They do great work, but they’ve learned that somehow nothing discussed in these meetings really matters, unless it does, at which point their ears immediately perk up. For example, if there’s a shift around RTO.

For the most part, the meeting sounds like “blah blah blah” white noise to them.

Purists

Then you have the purists who seethe about it not being a strategy.

“This is bullshit!” they say.

“This is a bad strategy! I read the book!”

“These people are incompetent, self-serving, or both!”

Confused Coherence Checkers

And then you have the folks who are genuinely confused. They aren’t naive like the purists. They know this is a bit of a show. But they are sitting there wondering if they somehow didn’t get the memo.

Given everything at stake, why NOT put a real strategy up there? Even if it were a bit watered down?

“Am I a chump for not playing the game the #2s are playing?”

“I mean, surely they don’t believe this is a strategy. Or do they?”

They feel tempted to raise their hand and ask a question, but immediately sense that might be a bad idea.

“This is a $150,000 meeting,” they wonder. “Am I missing something?”

They feel for the teams that are going to bear the brunt of the ambiguity. They aren’t just noticing that the strategy is incoherent. They are already imagining the meetings that will follow, the conflicting interpretations teams will make, the work that will continue despite the supposed priorities, and the people who may later be blamed for failing to execute.

“I feel that, with just a bit more work, we could have turned that into a great presentation!”

The last time they said that out loud, they were told to be more of a team player. They are also trying to calculate whether saying any of this now would help, embarrass someone, derail the meeting, damage their standing, or simply mark them as difficult.

They know that every presentation leaves things out. They know leaders cannot explain every constraint, tradeoff, political reality, or unresolved disagreement to the entire company. They are not asking for perfect transparency. What bothers them is that the gap between the presentation and reality feels large enough to change what people will do.

2x2: See It, Tolerate It

Imagine a 2x2.

• Cannot see it, Can see it

• Can tolerate it, Cannot tolerate it

The less experienced Believers can tolerate the incoherence, but they also can’t see it.

The Players can see the incoherence, but they can tolerate it. They might even relish it.

The Coherence Checkers can see the incoherence, and it bothers them. But oddly they are “in their element” when untangling it.

“Do we have to get into all the gory details? No. But we should at least help the team navigate the current reality and do our best to inform decisions.”

They are trying to reconcile what leadership says, what leadership believes, what leadership is doing, what employees will infer, what teams will try, and whether speaking up would damage their personal situation or the team’s situation.

The “But it isn’t a strategy!” purists can’t tolerate it, but they also don’t fully see it. They dismiss the social function of the performance entirely and are naive in their own right, though technically correct.

The checked-out people sit outside the matrix. They have figured out that neither seeing nor tolerating it matters enough to waste their energy on.

If these were animals, you’d have:

1. The Golden Retrievers: Slides! Go!

2. The Foxes: Some predators. Some socially intelligent. Some both.

3. The House Cats: Conserve energy.

4. The Terriers: Conceptual violation registered. I will not let it go.

5. The Border Collies: I’m watching. I’m getting more agitated. The signals don’t match. I can see the people. I can see the performance. I see the constraints. I can see what the teams are about to misunderstand. I’m trying to stop herding everyone.

The Border Collie Faustian Bargain

It is a Faustian bargain of sorts for the Border Collies.

You can become a Fox, and maybe have more impact, if you surrender your right to challenge the fiction. But you risk, over time, becoming responsible for persuading everyone else that it is true.

You stop asking whether the ambiguity is helping the organization and start treating its preservation as part of your job.

At first, you rewrite the confusing message for your team.

Later, you explain why leadership had no choice but to make it confusing.

Eventually, you find yourself correcting the person who points out that it is confusing.

The bargain rarely presents itself as, “Sell your soul.”

It is more like:

• “Be realistic.”

• “Don’t undermine the team.”

• “You’ll understand when you’re more senior.”

• “Get some influence first, then you can change things.”

The system teaches people that being “responsible” means protecting your boss, avoiding visible failure, keeping your options open, toeing the line, and making difficult outcomes and decisions appear impersonal.

The risk?

You begin producing the conditions you originally hoped to navigate. You tell yourself you are staying close to power so you can make things better, but your job gradually becomes about protecting the ambiguity, conveying it downward, and disciplining the people who point it out.

You become a Trapped Fox. You say, “I’m done with this place,” and then stay for six more years.

Or you become a Nomadic Fox, moving from company to company, arriving with fanfare and a reputation as a fixer or seasoned operator. With a flick of the tail, they locate the real decision-makers, learn the game, build a coalition, and become indispensable.

1. Enter as an outsider.

2. Decode the system.

3. Become indispensable.

4. Hit the limits of influence.

5. Decide the organization is hopeless.

6. Leave for a new, more interesting puzzle.

Back to the Border Collies

Back to the Border Collies, assuming they haven’t gone the Fox route (yet).

Of all the players, they are carrying the most in their heads.

Do you bark and name the contradiction in the room?

Backchannel it to someone who may actually have the influence to change something?

Herd locally by translating the ambiguity into something your team can use?

Leave the field after deciding that the organization’s coherence is no longer your responsibility?

Or conserve energy like a cat?

The Question:

What advice do you give Border Collies?

Comments appreciated!

Early in my product career, I wore many hats. I conducted research, analyzed data, sketched interfaces and wireframes, wrote requirements, coordinated releases, answered support questions, and even, shudder, wrote some code. Some of that was empowering. Some of it was amateur hour. But it was all bundled into the loose idea of “product management.”

Over time, those capabilities separated into dedicated roles and disciplines.

Designers, analysts, product marketers, program managers, and product-operations teams developed deeper expertise than any one generalist could reasonably maintain. Some joined cross-functional teams and became part of the daily work. Others became centralized experts that teams accessed through requests, reviews, and shared services. Repeatable parts of their expertise were then encoded into analytics platforms, design systems, research repositories, templates, standards, and workflows.

A role is only a temporary bundle inside the system. Capabilities themselves are distributed across people, teams, routines, tools, platforms, institutions, and external partners. When technology changes work, it rarely swaps cleanly for one person or role. Instead, capabilities move through several different motions:

• Specialization: broad work splits into deeper expertise.

• Diffusion: expertise spreads to more people.

• Centralization: expertise concentrates somewhere else.

• Integration: separate capabilities are coordinated.

• Embedding: expertise moves into tools and routines.

• Rebundling: separated capabilities come back together.

• Externalization: capability moves outside the organization.

• Elimination: work genuinely disappears.

• Loss: work disappears before its underlying capability has been reproduced elsewhere.

The Headlines

Scan the recent headlines and work appears to be changing in every direction at once.

Product managers are becoming “AI builders.”

Designers are becoming “designer engineers.”

Software engineers are becoming managers of AI-generated code.

Data platforms are absorbing parts of data engineering, analysis, and model development.

Middle managers are becoming “player-coaches” and “org leads.”

Product titles are collapsing into a single category: “builder.”

Roles are simultaneously broadening, disappearing, becoming more specialized, and merging with adjacent roles. There are several different motions happening at once. You have all the patterns I described above.

• Technical and analytical skills are diffusing across role boundaries.

• Product, design, engineering, and management tasks are being rebundled into broader roles.

• Knowledge is being embedded in AI tools and externalized to platforms.

• New specialties are appearing around architecture, evaluation, orchestration, reliability, and governance.

• Some repetitive work is genuinely going away. And some organizations (though they may not realize it yet) are losing the apprenticeship, coordination, context, and judgment that the old work helped sustain.

The headlines are observing different capabilities moving in different directions and collapsing that story into catchy headlines.

Capability Moves Through Loops, Not Stages

It is easy to imagine the motions in the design example as a linear journey, or a single, repeating loop. In practice, there are many loops occurring simultaneously, and each motion is creating pressure and tension for other motions.

Each motion changes the conditions around a capability and creates pressure for another motion. Specialization produces depth but ALSO handoffs. Diffusion creates access can ALSO eventually produce inconsistency. Centralization produces consistency but pulls decision-making away from the people closest to the work (or context away from centralized capability). Moving expertise into tools so people don’t have to perform the work directly (embedding) can weaken the learning pathways through which they develop judgment.

These loops do not occur in a fixed sequence. Several may operate simultaneously, at different levels and over different periods of time. A capability might be:

1. Diffusing across teams while control over its infrastructure centralizes;

2. Embedded in technology while new specialist work emerges at its frontier;

3. Externally supplied while internal capability is rebuilt around direction and assessment;

4. Rebundled into broader roles while its underlying knowledge becomes increasingly specialized;

5. Made more productive while the resulting increase in demand creates more work overall.

Each loop describes a recurring tension that may produce useful adaptation, harmful instability, or both.

What Actually Changes With AI?

I often ask myself, “Does AI change the physics of this pattern?” That is difficult to answer because a system can continue to obey the same underlying principles while the variables change so dramatically that those principles appear in entirely new forms.

There is already evidence that AI is affecting specialization, diffusion, centralization, integration, embedding, externalization, convergence, and loss. The real questions are to what extent, under what conditions, and with what second- and third-order effects. For example:

Diffusion

To what extent will AI allow experts to distribute their expertise more easily, without being constrained by the rigid interfaces we rely on today?

AI can make codified expertise available to more people, but we do not yet know how well it transfers tacit, contextual, or embodied expertise. Polanyi (1966) put it simply: we know more than we can tell. If that remains true, diffusion through AI will have a ceiling that is difficult to see from outside the expertise.

Integration

Can it actually lower the cost of specialists working with generalists and with each other?

The translation, handoff, and queueing burden created by specialization is real, and AI can friction-bust. But it hasn’t eliminated the “social” work of coordination, negotiation, and shared judgment.

Convergence and Rebundling

There’s a version of the future where individuals and teams recover capabilities that previously required several dedicated roles. A new kind of higher-level generalist.

AI does let people perform tasks across more role boundaries. But producing more kinds of artifacts is not the same as possessing the underlying expertise.

Specialization

And where will the opposite happen? AI is already creating new areas of specialization, new forms of technical complexity, and new specialists required to build, evaluate, govern, and adapt these systems. It may reduce demand for some specialist tasks while creating entirely new specialties around context, evaluation, orchestration, governance, and exception handling.

Embedding and Extension

Expertise is increasingly being encoded directly in tools and workflows, allowing people to act competently without possessing the underlying skill themselves. This is not hypothetical. It’s happening now. The question is what happens to the role that remains: it shifts toward direction, assessment, and oversight, which are genuinely different skills.

Centralization

AI may broaden local access to a capability while concentrating control over models, infrastructure, standards, data, and defaults in a small number of firms. More people become locally capable. Organizations become collectively dependent on increasingly centralized systems. Both things are true at once.

Loss

What happens when tasks disappear before the judgment, context, relationships, and learning pathways they sustained have been reproduced elsewhere? An organization may preserve output while losing the mechanisms through which people become capable of evaluating, correcting, and extending that output. The MIT Work of the Future task force (Autor, Mindell, and Reynolds, 2020) warned of precisely this: cutting out middle tasks can destroy workers’ skills, networks, and career pathways even when the immediate output looks intact.

All of these dynamics are in effect. What remains unknown is how far each motion will go, which will dominate in different contexts, and where those capabilities will eventually settle.

Limitations

What’s getting in the way of having honest discussions about AI’s long-term effects? The hype is deafening. Science always has trouble with longer-term effects. We don’t understand each other’s work well enough. And time plays tricks on how we process what is happening.

Hype

The hype is so omnipresent that many organizations are spending more time copying other companies, listening to vendors, and signaling that they are “doing AI” than conducting disciplined experiments around specific problems.

That is one of the defining dynamics of a bubble: spending money and taking visible action become signals in themselves. Each company’s activity becomes evidence that other companies should act, regardless of how much value is actually being produced.

This makes the underlying phenomenon harder to study. We are observing the technology, vendor incentives, executive expectations, labor-market anxiety, competitive signaling, and FOMO interacting at the same time. The hype does not mean the technology lacks value. It means adoption is not clean evidence of value.

Research

The second limitation is not unique to AI. Researching any change that touches individuals, teams, technologies, incentives, workflows, institutions, and markets creates severe limits on the general claims we can make.

A study may show that AI improves performance on a particular task (and population, model, and context). That does not necessarily tell us how a role will change, how a team will reorganize, what new coordination costs will emerge, or whether the initial gains will persist.

Many of the outcomes that matter most are also difficult to observe in short-term studies. Do people continue developing expertise? Will teams become more or less capable? Will organizations become more resilient, or more dependent? Strong evidence about a bounded task can easily be stretched into a much larger claim about jobs, organizations, or society.

Consider something like a codebase. Understanding how AI changes quality, maintainability, and the team’s collective knowledge would require a long-term, deeply contextual study. Long-term ethnographic studies of a codebase are difficult.

Flawed Mental Models

Beyond the hype and the limits of research, we bump up against a common foe: ignorance.

What we do know is that flawed mental models will remain flawed mental models, and they will continue to produce flawed assumptions.

1. If you imagine that making products is akin to operating a factory, you will likely miss the mark when assessing AI’s impact on product development. You will be more likely to assume that producing requirements, designs, code, and releases faster necessarily means producing better products faster.

2. If you miss the fact that data scientists spend a great deal of time helping people frame useful questions, you may assume that natural-language querying eliminates the need for data expertise.

3. If you treat a codebase as a pile of code rather than the cumulative record of decisions, constraints, failures, compromises, and hard-won knowledge contributed by everyone who built it over time, you may assume that generating replacement code is equivalent to replacing the system.

The average patient cannot determine whether a doctor truly understands cardiology, so they rely on licensing, training, certification, hospital privileges, and peer review to make that judgment for them. Inside companies, specialist work is rarely protected by an equivalent system.

Instead, people encounter familiar-looking outputs: a screen, a chart, an interview, a roadmap, a status report. Because the artifact looks understandable, they assume the underlying work is understandable too. The question shifts from “Is this person qualified?” to “Doesn’t this look like something I could do?” AI makes that confusion even easier. It can reproduce the visible artifacts of expertise long before it can reliably reproduce the judgment, context, coordination, and exception handling that made those artifacts valuable.

Time

These limitations share a common theme:

We have not had enough time.

History helps us recognize broad patterns, but not how they will settle around AI. We may be late in the S-curve of the current model architecture, early in the curve of AI adoption, and barely beginning the curve of organizational adaptation. AI is not one curve, but a stack of technical, economic, and social curves unfolding at different speeds.

So asking whether AI is “early” or “late” is too simplistic. The timing, magnitude, and interaction of these curves will determine which capabilities diffuse, which concentrate, what new organizational forms emerge, and how much damage the surrounding bubble dynamics cause along the way.

Conclusion

How do you introduce these ideas safely into your organization?

Carefully.

Raise a measured concern about the erosion of skill and judgment when AI use veers more toward substitution and delegation than augmentation and learning. This concern is already fairly widespread and accepted, so you should not get into too much trouble.

You might ask:

Which routine activities are helping people build the judgment they will later need for difficult situations?

And then:

If those activities disappear, how will that judgment be developed and maintained?

From there, introduce the capability motions and feedback loops. Do not present them as a grand framework that everyone must adopt. Use them as language for describing what people are already seeing.

• Where is expertise diffusing?

• Where is it becoming embedded?

• What is centralizing?

• What is being externalized?

• What has been rebundled into broader roles?

• Where are handoffs shrinking, and where are new ones appearing?

• Which loops seem to be producing the outcomes you want?

The point is not to classify everything perfectly. It is to make more of the system visible. The discussion about AI magic has been especially fruitful for me and several colleagues. Once people stop treating the technology as an independent intelligence arriving to join the workforce, they can begin asking more grounded questions about the system around it:

• What context does it have?

• Whose knowledge has been encoded into it?

• What remains outside its reach?

• Who is capable of assessing its output?

• Who controls the infrastructure and defaults?

• What happens when it reaches an exception?

Good luck! Whatever we can do to encourage richer conversations, grounded in reality, not mired in the weird duality of “humans vs AI”, will help us adapt and learn.

Here’s a simple model for thinking about AI success in organizations.

The Variables

Machine (Technology) Understanding

How AI actually works. The inputs and outputs. The failure modes. You don’t need an expert-level sense of the math, but you need a viable mental model of how the technology and what it is doing. Understanding the costs (based on a very incomplete current sense of the costs). Said another way, you have to get past the “magic” of AI.

Problem Understanding

Context, domain experience, skill, customer knowledge, constraints, history, politics. Everything that makes your problem a real-world problem and not a generic problem. If you are at risk of falling into the Build Trap, you’ll just do it faster with AI.

Practice Evolution

How the work actually gets done. Experimentation, feedback loops, workflow adaptation, team learning. The connective tissue between knowing and doing. Accepting that there are no “best practices” yet. Things change every week.

Social Contract

The organizational commitment that improvement gains will be used in line with a shared agreement, not against the people who generated them. And that the work remains compatible with who people are and who they are becoming.

Resistance to AI mandates is a perfectly rational behavior. Kaizen only works when there’s a credible social contract that the gains from improvement will not be captured against the improvers. Without that contract, you don’t get resistance out of ignorance. You get resistance out of rational self-preservation.

But it is not just defense. This is not just people “afraid for their jobs.”

The jury is genuinely still out. The data is not there (especially over any reasonable period of time). The resistance is not merely emotional or self-interested. In many cases it reflects an accurate reading of the evidence. People are being asked to cooperate with a premise that has not been proven, and that might be used against them if it ever is.

Failure Modes

The formula is multiplicative. Any term at zero in the numerator kills the product:

• Machine × Problem × No Practice Evolution
  You know what AI can do but never update how you use it. You assume the practices have been figured out.

• Machine × No Problem Understanding × Practice
  Fast iteration in the wrong direction. No ability to evaluate the results.

• No Machine Understanding × Problem × Practice
  Deep expertise, no sense of what the machine is doing underneath. When the technology changes you are unaware. Can slip into “it is magic” thinking.

• Machine × No Problem Understanding × No Practice Evolution
  Demos and pilots disconnected from real work. No practical way to put it in motion. No real applicability.

• No Machine Understanding × Problem × No Practice Evolution
  You know it isn’t working, but have no idea why. And no way to refine your practices.

• No Machine Understanding × No Problem Understanding
  New workflows for their own sake.

And when the denominator is broken—when people don’t trust where the gains will go, or feel the work threatens their identity—the formula collapses regardless of how strong the numerator is. You get rational resistance, not incompetence.

Most AI mandates fail because they try to force one term in the numerator high while ignoring the other two. “Everyone must use AI” is a Practice Evolution mandate that assumes Machine Understanding and Problem Understanding will take care of themselves. They won’t.

And even if the numerator is strong (even if people understand the machine, know the problem, and are ready to experiment) without a credible social contract, honest effort never arrives.

The denominator is not optional. It is the precondition for everything above the line.

AI does not eliminate complexity, coupling, coordination, uncertainty, or decay. But it can change the cost, speed, and feasibility of responding to them.

or:

AI does not work in the same way across different portfolio puzzles.

This is absolutely not a new idea. Companies have long thought about software as a portfolio of assets, investments, bets, and liabilities. What interests me here is not the metaphor itself, but what happens when we combine it with value, carrying cost, movement over time, and AI.

This is more of a Sunday-morning diagramming and thought-provocation post than a fully worked-out framework. None of the underlying ideas are especially new. I’m mostly interested in placing a few familiar concepts next to one another and seeing what becomes easier to notice.

Let’s go!

Carrying Costs and Value

Some assets create substantial value and remain relatively cheap to carry. Some create value but have become expensive to change, operate, and support. Some are still bets. Others continue consuming resources long after their economic case has weakened.

The question is not whether the code is “good” or “bad.” It is:

What value does this asset create, what does it cost us to carry, and what should we do next?

That gives us a simple 2×2.

Place each software asset according to the value it creates and what it costs to carry. Its position suggests the basic economic posture: incubate, compound, refinance, or liquidate.

Motion and Arc

Consider the arc of a startup.

It begins with low value and low carrying costs: a relatively cheap bet. Some bets prove their value and move toward Compound. As they scale and mature, carrying costs rise, pushing them toward Refinance. From there, a healthy company either restores the economics or eventually Liquidates the asset when its remaining value no longer justifies the cost of carrying it.

Other bets never prove much value at all. When the team keeps building anyway, the product moves right without moving up, turning a cheap experiment into an expensive obligation: the build trap. Liquidation is not necessarily failure. Refusing to liquidate often is.

Once you see the portfolio this way, you can also see the different moves available.

You can expand returns, reduce drag, harvest an aging asset, scale with drag, or slip into negative compounding. But the “pure” movements are mostly abstractions. Products rarely move perfectly up, down, left, or right. Growth usually brings some additional drag. Reducing carrying costs may sacrifice some value. Harvesting is often a deliberate movement down and left.

The more interesting question is not simply where an asset sits. It is:

Which direction is it moving, how quickly, and is that movement intentional?

Debt

These categories and movements make technical debt easier to see in financial terms. Debt is a financing choice: we accept higher future carrying costs in exchange for reaching value sooner.

• In Incubate, that can be dangerous because we may take on obligations before proving the asset has value.

• In Compound, it can be rational leverage if the added drag helps expand returns.

• When the interest begins to consume too much of those returns, the asset moves toward Refinance, where the task is to restructure the system and restore its economics.

• And when the asset no longer creates enough value to justify either the debt or the cost of repaying it, the right move may be Liquidate.

The same technical debt can therefore be productive leverage, an expensive constraint, or a liability, depending on the value of the asset carrying it.Portfolio Over Time

Platforms

These categories also help clarify the difference between an effective and ineffective platform strategy. A good platform may create a visible concentration of cost, but it earns that investment by helping many other assets increase value, reduce carrying costs, or do both. A bad platform becomes another high-cost asset that pushes complexity, dependencies, and coordination overhead across the portfolio. The relevant question is not how much the platform ships or how many teams use it. It is whether the platform improves the economics of the assets that depend on it.

We’ll get into more traps later, but platforms have a fun assortment of traps to contend with:

• False Economies of Scale: We centralize similar-looking work without proving that sharing it will reduce total carrying cost.

• Platform Too Early: We standardize an emerging pattern before we understand what is actually stable and reusable.

• Platform Too Late: We wait until duplication and divergence make convergence painfully expensive.

• Adoption Tax: We reduce complexity for the platform team by pushing integration, migration, and exception handling onto consumers.

• Mandate Trap: We mistake compulsory usage for evidence that the platform improves anyone’s economics.

• Dependency Bottleneck: We centralize capabilities and accidentally centralize waiting, prioritization, and failure.

An Intentional Portfolio

There are always inflows and outflows between these categories over time. Some movements are intentional portfolio choices. Some are unavoidable byproducts of growth, scale, and aging. Others happen accidentally, through oversight, avoidance, or outright negligence.

So portfolio health is not only about the mix of assets at a given moment. It is also about what is flowing where, how quickly, and whether anyone is managing the movement.

This frame also makes it easier to see familiar organizational traps as recurring patterns of movement between the categories.

The Traps

The Build Trap is what happens when Incubate fails to become Compound, but keeps accumulating carrying cost. Limits to Growth appears when Compound succeeds, then steadily gives way to Refinance. The Success Trap emerges when current winners dominate so completely that the next generation of Incubate assets disappears.

There are plenty of other patterns hiding in the diagram. Some are failures to prove value. Some are failures to preserve the economics of value. Others are failures to renew, simplify, or finally let go.

Not A Monolithic Situation

The critical question when it comes to AI is:

Which movement are we asking AI to accelerate?

AI does not enter a neutral system. It enters a portfolio with a particular mix of bets, winners, drag, and aging assets. Depending on the situation, it can help move an asset upward, leftward, or back into healthier economics. It can also accelerate exactly the wrong motion.

Depending on the answer to that question, you can better understand the risks you need to mitigate.

For example:

• In a Build Trap, AI can help teams prototype and synthesize feedback faster. It can also make it much easier to keep building without stronger evidence of value.

• In Limits to Growth, AI can help with refactoring, migrations, dependency analysis, and automation. It can also pour more throughput into the very constraint that is already slowing the system down.

• In a Success Trap, AI can make adjacent exploration much cheaper. It can also help the company optimize its current winner so aggressively that it stops looking for what comes next.

• In a Zombie Portfolio, AI may finally make old systems affordable to retire. Or it may make them cheap enough to tolerate forever.

The same capability can improve the economics of the portfolio or deepen the pattern that created the problem. So the AI question is not simply:

Will this make us faster?

It is:

Faster in which direction?

AI does not eliminate complexity, coupling, coordination, uncertainty, or decay. But it can change the cost, speed, and feasibility of responding to them.

The physics do not change. The coefficients, thresholds, and economics may.

• How to thrive when you feel misunderstood at work, or

• How to thrive at work when you think too much

So here’s the introduction, along with a bunch of posts that tie in. A favor: if any of this resonates, and you have someone in mind who might benefit from reading it, please forward it along.

Subscribe now

Introduction

Have you ever heard something like this at work and felt hurt, confused, angry, defensive, or even combative?

“You’re always overthinking things.”

“You see misalignment that isn’t there.”

“You make things too complicated.”

“You are too theoretical.”

“You have to learn how to keep it simple.”

“You need to keep your emotions under wraps.”

“You focus on problems and not solutions.”

“You keep bringing up things that are settled, why?”

“You spend too much time analyzing things.”

“You’re not comfortable with ambiguity and uncertainty.”

“You can’t push back like that in front of the team.”

The Garden

Feedback like this can create the visceral response we all know from early childhood onwards: “You just don’t understand!” At work, as an adult, the words change. “I’m perfectly comfortable with uncertainty... you’re the one trying to gloss it over!” But the feeling is the same. Misunderstood.

“You don’t get me!”

It can be especially difficult when these are the very things you value: going deep, noticing what doesn’t add up, sitting with something until it makes sense. Some people value that too, and they have let you know. You know it can be a strength. Hearing it framed as a flaw still lands like an attack on who you are. At a minimum, it would be nice if the feedback were balanced. But in the rush of day-to-day work, rushed 1:1s, and too many things competing for attention, it rarely is.

A memory from childhood. My family redesigned our garden. They loved gardens. A columnist from the Washington Post came to our house with an illustrator and published a story with me in it, in the garden, like a children’s book. A fantasy. I was there, curious, daydreaming, exploring, wondering. The illustrator sketched question marks around me.

The real image is fading in a box somewhere. But the mental image is lodged in my brain like it was yesterday.

I know this sounds like an odd thing to bring up. But whenever I got that kind of feedback at work, something in me would flash back to that image. The garden. The celebrated question marks. The feeling of being celebrated for curiosity, for wondering, for not knowing yet. And then the fluorescent-lit room (or Zoom) where someone is telling me I’m too much of exactly that. The dissonance was almost physical. Like two versions of the world that couldn’t both be true. One where wondering is who you are and people delight in it. Another where it’s the thing you need to fix.

We all have that version of the garden that we want to protect and return to at will as we navigate the adult world.

Thriving, Work, and Self-Worth

For years I treated every piece of feedback like this as a verdict.

One story: I’m the problem. Too complicated, too emotional, too abstract and theoretical, too easy to read. The other: the place is broken, the leader is careless, the “system” isn’t working, and it’s on me to say so. Neither story, or the countless sub-plots that emerged, helped. The question that actually mattered was harder and more practical: how much of this is mine to work on, how much is just the game, and can I tell the difference before it eats me alive and I do/say something that I’ll regret.

Put another way: What path do I take to thrive?

It’s always important to process feedback. If you’ve gotten the same kind of feedback repeatedly, from different people or in different contexts, that’s probably something worth paying attention to. Paying attention might mean changing your behavior somehow.

It might also mean detachment, seeking out new environments where your strengths are a good fit, or deciding this isn’t a fight worth having on those terms. It might even mean recognizing that you’re drawn to situations that aren’t healthy for you, where you’ll hear the same feedback again and again, because you respond to it the same way each time. Psychologists call that a repetition compulsion: returning to a familiar pattern even when it hurts.

But what took me too long to learn: you still have to disconnect your sense of self-worth from the workplace. In fact, not just “even if” you love what you do, but if. If you love what you do and want to keep doing it, you’ll need to figure out how to detach from using the workplace as a proving ground for your worth. Be brutal about your craft. Be your own hardest critic. Seek out people who will push you, challenge your thinking, and refuse to let you coast. But that’s the work, not the workplace. The work deserves your full intensity. The workplace doesn’t.

That’s not the same as ignoring what you hear. It’s recognizing something unique about work, something organizational researchers have documented for over a century.

Max Weber showed that modern companies are built around roles, not whole people. Ferdinand Tönnies contrasted organic community with instrumental association. Arlie Hochschild studied emotional labor, the requirement to perform feelings as part of the job. The through-line is consistent: the workplace is not a proving ground for you and your value. It can feel like one. Feedback lands like a verdict. You tie your identity to fixing what’s broken, or to being the person who “goes deep” or “calls out BS.” But a company is a different kind of place than the parts of life where belonging and recognition actually work that way.

I joke that your average pirate ship was more of a community than your average company. But it's true. Pirates elected their captains. They voted on destinations. They split plunder by agreed-upon shares and could depose a leader who lost their trust. The crew's belonging wasn't contingent on quarterly results or a reorg. It was bound by shared risk, mutual dependence, and a code they wrote together.

Your company, probably, has none of that.

You’re replaceable by design and belonging is conditional. The org can pivot, reorg, or lay you off in ways that have little to do with your insight, your effort, or your character. None of that means you don’t matter. It means the workplace is a terrible instrument for measuring whether you do. When your self-worth is hooked to that instrument, every misunderstood comment becomes existential.

The skill isn’t caring less. It’s caring differently: about your work, your growth, your standards, your relationships, without making the org the mirror that tells you if you’re OK. Only then can you start to sort out what’s yours to work on from what’s just the game.

The Game. The Work

So back to the two questions: how much of this is mine to work on? And how much is just the game?

The Game

Start with the game, because a lot of people pretend it isn’t there. Nowhere is that pretense stronger than in the advice to “bring your authentic self to work.”

Jeffrey Pfeffer has shown that getting ahead in organizations is closely linked to self-promotion, political skill, and a willingness to make others uncomfortable. Not with being right. Not with being helpful.

Erving Goffman’s work helps explain the gap. In social life we are always, to some degree, on stage. There is a front stage, where we perform a version of ourselves for an audience, and a back stage, where we drop the act. Work is mostly front stage. You are not just sharing what you think. You are performing a credible professional self. When feedback says you’re “too theoretical” or “not a team player,” Goffman would say your performance didn’t match what the scene required. That may have nothing to do with whether you were right. You brought back-stage honesty into a front-stage moment that demanded confidence and alignment.

Venkatesh Rao, not a scientist but a wonderfully sharp observer of organizational life, took this further with what he called the Gervais Principle. His framing: “Sociopaths” see the game clearly and use it (they advance). “Clueless” genuinely don’t see it (they stay cheerfully in the middle). And “Losers” see it, are bothered by it, but make a conscious bargain to endure it anyway. If you’re reading this post, you probably see it. The question before you is whether to make the bargain, refuse it, or find some fourth path no one has handed you yet.

More research:

• Organizations disproportionately advance people with psychopathic and narcissistic traits, even when those traits undermine team performance (Babiak & Hare; Babiak, Neumann & Hare, 2010).

• Political skill predicts career advancement more reliably than job performance alone (Ferris et al.; Munyon et al. meta-analysis).

• Overconfident individuals attain higher social status, even when peers can see the overconfidence (Anderson, Brion, Moore & Kennedy, 2012).

• At elite firms, cultural similarity and “fit” outweigh technical competence in hiring decisions (Rivera, Pedigree).

• Narcissistic traits predict leader emergence but not leader effectiveness. The traits that get you the role aren’t the traits that make you good at it (Judge, Piccolo & Kosalka, 2009).

That is my pushback on “bring your authentic self to work.” The core issue is bringing authenticity and the workplace into the same conversation, as if they were built for each other. Authenticity matters. It does. Workplaces matter too. A job is nice, and can be incredibly rewarding. Your work can matter deeply to you. None of that is what I’m arguing against. But when you merge the two, when you treat the org as the place that validates your real self, you set yourself up for exactly this kind of hurt. Companies ask for performance all the time, including the performance of authenticity, and then punish people when their actual inner life doesn’t fit the scene.

Hochschild called it emotional labor. We might also call it a trap: be yourself, but not like that.

Company Vs. Community

I have a theory that it helps to imagine your workplace as a literal stage. A Shakespearean tragicomedy. People enter, perform their roles, navigate alliances and misunderstandings, deliver monologues that land or don’t, and exit. Some scenes are brilliant. Some are painful. The drama is real, the stakes feel real, but it is still a production. Nobody is their role, even when they forget that.

Contrast this with a community: a neighborhood, a family, a long friendship, a religious congregation. Yes, we perform in communities too. Goffman would be the first to say so. The difference isn’t performance versus no performance; instead, it’s about who holds the script, what’s at stake when you flip the script, and whether belonging survives you revealing some deeper level of yourself.

In a close relationship, dropping the mask is how intimacy works. At work, dropping the mask is often how you get the feedback at the top of this post. Hochschild drew the line here: we all do emotion work in private life, but emotional labor is when the performance is sold for a wage and managed by someone else. Tönnies called the community end Gemeinschaft: organic, embedded, durable. Not because communities are pure or free of conflict. Because belonging there isn’t contingent on your output.

The workplace is Gesellschaft: instrumental, contractual, built around functions. The cast rotates. The script changes with every reorg. Your part can be written out for reasons that have nothing to do with your talent or your character. Academics distinguish these two structures because they operate on fundamentally different logics. One is built on identity and belonging. The other is built on roles and exchange. When you confuse which one you’re in, feedback about your performance starts to feel like rejection from your tribe. It isn’t. It’s a note from a director about how you played the scene.

All to say: how much of the feedback you treat as “yours to work on” depends on what you actually want. You can’t sort out the feedback until you’re honest about which game you’re playing, or refusing to play.

Aside: I need to pause here. Because for some people, “performing” at work is not a strategic choice. It is the only way to survive. Psychologists studying neurodivergent masking have documented the sheer exhaustion of suppressing natural responses all day just to be read as competent. Researchers studying code-switching in Black professionals, women in male-dominated fields, queer people in conservative workplaces, immigrants navigating cultural expectations, have found the same thing: the performance isn’t optional, and the energy it consumes is enormous. The burden of navigating the front stage is not equally distributed. Some people are performing with a script written for them by default. Others are performing against a script that was never written with them in mind. “Just change your approach” lands very differently depending on how much you’re already spending just to be in the room.

Which brings us to…

The Work

OK. Now on to “how much of this is mine to work on?”

I have a great friend who once said something that was both incredibly funny (if you knew them), and very poignant. “I reliably rub certain types of people who believe I’m a certain type of person, the wrong way.”

I could relate immediately.

There’s so much packed into that sentence. Self-awareness. The wry acceptance. The recognition that the friction isn’t random. Certain types of people. A certain belief about who you are. Also: they’re typecasting the people who typecast them. They know it. That’s part of what makes it funny.

That’s what makes it hard, too. Because if the pattern is reliable, you have to ask: is this a me thing, a them thing, or a fit thing? Probably all three. The version of you they’re reacting to may not be who you actually are. But the fact that it keeps happening? That’s data pointing to something...

Here’s the wisdom I keep coming back to.

It doesn’t matter what “type” you think you are. It doesn’t matter what type someone thinks you are. It doesn’t matter how you categorize people: Big Five, the Gervais Principle, “the spectrum,” high agency versus low agency, systems thinker versus executor. It doesn’t matter how your worldviews stack up, or whose framework is more correct, or who “gets it.” If what you’re doing isn’t having the effect you care about, you’ll need to change something. Not because you’re wrong. Not because they’re right. Because the effect is the thing, and the effect is what you can actually work with.

In many ways this is incredibly liberating. William James, the father of pragmatism, argued something similar: stop asking what category something belongs to and start asking what it does, what effects it produces. You are not your type. You are a series of choices and their consequences. If the consequences aren’t what you want, you can make a different choice. No identity crisis required.

There’s a psychological framework that operationalizes this beautifully. Marsha Linehan’s Dialectical Behavior Therapy includes a skill she calls “effectiveness.” The instruction is deceptively simple: focus on what works. Let go of being right. Act as skillfully as you can to meet your goals in this situation. Not “pretend it doesn’t hurt.” Not “they were right all along.” Just: given what you actually want, what’s the move that gets you there? The feelings can be there. The history can be there. The garden and the question marks can be there. But the next action doesn’t have to come from whatever wounds you are carrying; it can come from your goal.

The elephant in the room here is the effort, the potential trauma, the psychic pain that accompanies all of this redirection and masking. It can feel deeply unfair, especially when you’re putting in all this effort and the people giving the feedback seem to be just comfortably winging it.

The cognitive load is real. But the research offers something useful: the load isn’t fixed. It depends on how you regulate, not just that you have to.

Some Strategies

James Gross’s emotion regulation research: reappraisal (changing how you interpret the situation) costs far less than suppression (masking your response). “This feedback is about the scene, not my worth” is cheaper than “don’t react, don’t react, don’t react.”

Ask yourself: What story am I telling myself about this feedback right now, and is there a less expensive interpretation that’s still honest?

Peter Gollwitzer’s implementation intentions: pre-decide with “if-then” plans so you’re not burning working memory deliberating under pressure. The decision is already made before you walk in.

Ask yourself: What’s the situation I keep finding myself in, and what do I want my default move to be next time?

Steven Hayes and ACT: trying not to feel something is more expensive than letting it be there without letting it drive your next move. Acceptance isn’t approval. It’s dropping the internal fight.

Ask yourself: What am I spending energy fighting internally that I could just let be there while I act anyway?

Deci and Ryan’s Self-Determination Theory: when an adaptation feels chosen rather than forced, it costs less. Find your own reason, not just theirs.

Ask yourself: Is there a reason to adjust my approach that makes sense to me, on my own terms, not just because someone told me to?

And sometimes the most effective move is structural: choose environments where less regulation is needed in the first place.

Ask yourself: Am I spending most of my energy on the work, or on surviving the room?

Here’s what it comes back to. If you love your work, if you enjoy the doing, that is the thing to protect. Not your reputation in a meeting. Not whether someone “gets” you. The visceral day-to-day connection with making something, solving something, building something. Or maybe it’s not even the work itself. Maybe it’s what the work makes possible: your family, your hobbies, the causes you care about, the life you’re building around it.

That’s what’s worth protecting. And disconnecting your sense of self-worth from the workplace is the key to getting the distance you need to protect it. You need enough distance to take feedback without crumbling, to play the scene without losing yourself, to notice what’s off without being consumed by it. Enough distance to stay in the game long enough for the things you actually care about to keep growing.

Here’s what it comes back to. Remember the garden.

Not the literal one from my childhood…your garden. The thing that drew you to this work in the first place. The curiosity, the making, the solving, the building. The feeling of being absorbed in something that matters to you. Maybe it’s the craft itself. Maybe it’s what the work makes possible: your family, your hobbies, the causes you care about, the life you’re building around it. Either way, there is something you’re tending. Something that grows when you give it attention and withers when you don’t.

If this resonated, here’s more I’ve written on these themes:

• 20 Things I’ve Learned as a Systems (Over) Thinker — The practical list: brain always overtime, pace your delivery, don’t show all your work, celebrate wins, know when to fold.

• How (And Why) We Help — When your instinct to help becomes self-sabotage. The tension between working within a system you don’t believe in and questioning it, and how childhood patterns show up in professional urgency.

• Maximizers vs. Focusers — Why the tension between “move fast, keep options open” and “go deep, stay coherent” keeps producing the same conflicts. Different defaults, not different intelligence.

• “You’re Making This Too Complicated” — Why people interpret your approach as “too complicated” or “not bringing solutions.” Different action-vs.-analysis defaults, and how to bridge them without losing yourself.

• Just and But — Bias toward action vs. bias toward questioning as complementary styles, not character flaws. How trust changes the dynamic.

• Probably, Just, Maybe, and Perhaps — Your default explanation when things go wrong, and how to practice generating alternatives before you act.

• Self-Gaslighting and the Doubt Loop — What happens when low-level dysfunction erodes your confidence from the inside. How to recognize the pattern and reconnect with what genuinely fulfills you.

• Why Reflexive Thinkers Need to Take More Breaks — The extra cognitive and emotional cost of holding multiple perspectives at once. Why you’re more drained than your peers and what to do about it.

• The Integrator Burden — The emotional labor of connecting teams when formal structures don’t recognize that work. Why underrepresented people often carry it, and what to do about it.

• The tough reality of being a “glue person” — Silent wins, public mistakes. Practical self-protection for people who hold things together and pay the cost of visibility without credit.

• Perception Asymmetries (and Your Career) — The gap between what your champions see and what external stakeholders perceive. Why competing narratives form, and how to calibrate.

• Working With Very Confident People (Who Dismiss You) — The threat response it triggers, the attribution bias it creates, and the practical move: pick battles strategically, protect your energy.

• The Self-Sealing Argument Trap — When feedback is structured so any response confirms the accusation. How to recognize it, name it, and stop playing a game you can’t win.

• In That Space Is Our Power — Frankl’s space between stimulus and response, applied to being dismissed at work. What you actually want when you feel unheard, and how to avoid escalation.

• That’s Not What They Are Hearing — Why conversations go circular when both people respond to the surface issue and miss the feeling underneath.

• No Unforced Errors — Why change agents get judged more harshly for off days. Negativity bias, halo/horns effect, and managing energy for the long haul.

There’s more, but it is getting late!

AI can be your scribe, your thought partner, data-stitcher, or your counselor.

And AI can be your consigliere, your spin doctor, or your propagandist.

Team superpowers.

Or Orwellian Ministry of Truth.

If a behavior is blocked by poor signal visibility, high workflow friction, missing scaffolding, weak shared representations, or limited procedural fluency, AI can help.

But if it is blocked by low voice safety, misaligned incentives, low legitimacy, controlled motivation, fear, shame, or learned helplessness, AI will not fix it by itself, and may make it worse by making the existing avoidance, politics, or performance theater easier to automate.

AI amplifies the direction you are already moving.

If an organization is on the cusp of improvement, with enough curiosity, trust, openness, and appetite for tackling/facing hard truths, AI can reduce friction and help that movement gather speed.

But if the inertia is toward avoidance, self-protection, green dashboards, and political storytelling, AI can accelerate that too.

If your team is swimming in menial data entry friction, spending countless hours on lightweight “copy then recontextualize” tasks, or fighting rigid data schemas with no room for nuance, then AI may remove real drag. Maybe AI unlocks a practice your team always wanted, but could never sustain because the clerical burden was too high. This is something I see at the day-job with our customers.

But if your organization has a track record of weaponizing data, using dashboards to corner PMs into “do this and this” conversations, treating metrics as ammunition rather than inquiry, or collecting information without using it to make clearer decisions, then AI will not magically make the system better. It will make it worse.

Context as hivemind. As an enabler to mētis (see James C. Scott for more on mētis, but in short: situated, practical, local knowledge that lives in experience, context, and judgment, not just formal rules or abstract models).

Context as control. An enabler to legibility from above: surveillance, standardization, simplification, and administrative power.

As my friend Summer Koide, a great coach, described it: AI can be like hydrogen peroxide, bubbling up to show you where the wound is…the hidden tensions, the weak signals, or the unresolved tradeoff.

Note: Summer coined AI-as-consigliere as well! Thank you!

Or AI can be the air freshener.

Kent Beck has described AI agents as less like loyal assistants and more like unpredictable genies: they grant wishes, but not always the wish you meant. The genies reply to requests with finger guns: “Yes boss!” “You got it, boss!” But, as in many classic genie stories, there is a dark side to that relationship. The genie grants the wish, but not always the intent behind the wish. It may do exactly what you asked while violating what you meant.

But there’s a twist.

The classic genie warning is that the genie gives you what you asked for, not what you meant. But there is another, darker version: the genie gives you exactly what you meant. If what you meant was “help me avoid this conflict,” “make this look green,” “turn this into a defensible narrative,” or “give me the evidence that supports the decision I already want to make,” AI may be very good at that.

If you are trying to coerce a coworker into doing something, navigate a performance discussion, or prove that your department is not the cause of X, AI will willingly comply.

“Tell my boss that they are wrong, but soften the message. Also, they’re a Libra and an Enneagram 6, so make it feel psychologically safe.” The output may sound humane, nuanced, and emotionally intelligent. But the underlying intent is manipulation, blame-shifting, or avoidance under a shroud of LLM-generated empathy.

Or the performance discussion. Here’s your “prompt”, and no one will stop you from doing this (trigger warning: workplace manipulation, bad-faith performance management, and termination dynamics):

I need help engineering a performance conversation that appears compassionate, fair, and fully aligned with company values, while making it very difficult for this employee to recover. The desired outcome is that they either self-select out or we create a clean, defensible record for termination. Do not make this sound punitive or predetermined. Use language about support, clarity, expectations, ownership, psychological safety, and growth.

Please help me:

• Frame subjective concerns as observable patterns.

• Use “stakeholder trust” and “role expectations” without sounding accusatory.

• Include just enough support language to make the process look sincere.

• Avoid anything that could imply bias, retaliation, or a decision already made.

• Make the employee feel the seriousness of the situation without giving them a clear point to argue against.

The tone should be warm, values-driven, and legally careful. The effect should be unmistakable: they should understand that staying will be exhausting and unlikely to succeed.

In theory, an employee could file a complaint or legal claim to challenge a bad-faith performance process and seek discovery of related records, but given that this has been happening forever, just with less polished tooling, would they be successful?

Could AI help a manager self-interrogate their own treatment of the employee, challenge their mental model, notice where they are over-attributing, and understand their report better? Absolutely.

A manager could ask: “Help me challenge my own story about this employee. What might I be missing, over-attributing, or contributing to? What evidence would change my mind?”

Will that happen?

Sometimes. But only if the manager asks for that. AI will not automatically turn a defensive prompt into a reflective one. If the manager asks for a cleaner case, it will help build the case. If the manager asks for a better understanding, it can help build that too.

This might sound like an “AI is just a tool” piece. And, sure, that is basically part of what I’m saying.

But “just a tool” has never quite sat right with me.

There is a personal responsibility to understand the tool you are using. To build the metacognition to notice your own intent. The empathy and “other-metacognition” to wonder what someone else is experiencing. And the computational metacognition to ask: what is this thing doing, smoothing, inferring, omitting, or making weirdly legible?

But it is also not just an individual responsibility story. We co-design our environments, and then our environments shape us back. At some point, when the patterns drift far enough away from any one person’s ability to intervene, it is fair to call them systemic.

So we need to be having discussions about all of it: individuals, systems, behaviors, etc.

And for those hoping AI gives us some kind of epistemic justice because it is computational, probabilistic, and “objective”... I think they are in for a surprise. Or maybe they know that already. Maybe what they mean is that AI might finally make their justice easier to operationalize.

Navigating this polarity is a challenge for our times.

It has always been a challenge: but now the challenge has been amped up.

What is your organization doing to get this right?

Or wrong (on purpose)?

Or wrong (out of ignorance)?