Why I pass on great teams

The best team I met this year pitched me in February. Two repeat founders, one of them out of a frontier lab, shipping weekly, revenue real and growing, references that glowed without being coached. The product did exactly what the demo promised, which I checked, because I always check. I spent two weeks trying to get to yes and then passed, slowly and unhappily, in a phone call I still think about.

The reason had nothing to do with them. That is the point of this essay, because the reason it took me two weeks is that everything about the team said yes, and team quality is the variable early-stage investors are trained to weight above everything else. I no longer think that weighting is right, at least not in AI, at least not now.

Execution became table stakes

The old logic was sound. Building was slow and hard, so a team that could really execute was rare, and rarity is where returns come from. Back a great team in a mediocre market and the team would find the better market eventually.

The cost of building has since collapsed, and it changed the arithmetic. A competent team stands up a credible product in weeks. The gap between the best teams and merely good ones, measured in what actually ships, is narrower than it has been in my investing life. Great execution still matters, but it has become the entry ticket rather than the edge. When most teams in a category can build the thing, the outcome is decided by something none of them control: the shape of the market they are standing in.

By shape I do not mean size. Every deck has a TAM slide and I have stopped reading them. Shape is four questions. Who is the buyer, and what is happening to them? Which budget line does the money come from, and is that line growing or dissolving? Where does value pool in the chain once the category matures? And what does the problem itself do under five more years of model progress? I have written about what happens to a product when the model underneath it gets ten times better. This is the same question asked one level up, about the market instead of the product.

Three shapes that make me pass

The shrinking denominator. The February team sold quality tooling into support operations: scoring, coaching, and managing large teams of human agents. Beautiful product, real pain, happy buyers. And every quarter, those same buyers were cutting the human seats the product billed against, using AI, sometimes using tools from the same conference booths. The company's own ecosystem was draining its market. They were rowing a very fast boat on a lake with the plug pulled. Best case, they execute perfectly and dominate a category on its way to being small. I could not write a check against that ceiling, and the ceiling was nothing they could out-build.

The pass-through market. Some markets exist as a gap between two platforms, and gaps close. The product is genuinely useful today because the workflow owner on one side has not bothered to build it yet. But value pools with whoever owns the workflow, and when the category proves itself, the workflow owner either builds the feature or buys the leader at a price that reflects a feature, not a company. You can win every logo in a market like this and still discover that winning the category was never the same thing as the category being worth winning.

The problem with a half-life. Some problems are not permanent features of the world. They are artifacts of what models cannot do yet. Entire products exist to manage a specific model weakness: to route around it, review it, patch it with process. The market is real, urgent, and well funded, and it is also evaporating on a schedule set by someone else's research lab. I met a team last autumn whose roadmap quietly depended on models staying bad at exactly the thing models were improving at fastest. A market can be real and temporary at the same time, and the deck cannot tell you which, because at day zero they look identical.

How I actually do the read

The mechanics are unglamorous. Before I decide, I write one page describing the market as I think the models will make it, three years out. Not the product, the market: who still has the problem, who owns the workflow by then, whose budget the money comes from, what the incumbents have absorbed. Then I ask the only question that matters: if this team executes perfectly the whole way, does the market they arrive in still pay for what they built?

Founders pitch the market as it is. That is not a flaw. They live at day zero, and the pain in front of them is real. But an early-stage check is underwriting the market as it is becoming, and those are different documents. The divergence between them is where most of my passes now come from.

There is a tell I trust in diligence. I ask the founder: what has to stay true about the models for this business to work? The strongest teams answer instantly and the answer makes them visibly uncomfortable, because they have done the read themselves and know exactly where the exposure is. The teams that answer smoothly, or answer that model progress only helps them, have usually not done the read at all. Discomfort, oddly, is the bullish signal. It means they are steering with their eyes open.

The cost, and keeping myself honest

This discipline is expensive and I want to be honest about that. Passing on excellent people means being wrong in public sometimes, because markets change shape. A platform stumbles, a buyer's budget migrates, a category that looked like a feature turns out to be a wedge. So every pass of this kind gets a written paragraph, dated: the shape I saw, and what would have to happen to prove me wrong. I review the file twice a year. It is the only way I know to distinguish between a read that was wrong and a shape that genuinely moved, and the two demand opposite corrections.

The other thing I do is tell the team the real reason, specifically, which almost nobody enjoys hearing and the best founders remember. I keep a list of every team I passed on purely for market shape, and it is my highest-priority inbound. Teams outlast markets. The February founders are already circling a different wedge in the same domain, one that gets stronger as the human seats disappear instead of weaker. If they land it, they will get the fastest yes I have ever given, and they know it.

I do not pass on great teams because I doubt them. I pass because I believe them. They will build exactly what they say they will build, and then a market that is dissolving underneath them will decline to pay for it. My job is to hold two reads apart: the team, and the lake it is rowing on. The team is the easy read. The lake is the job.