Wedge First, Platform Never

In about six weeks, a venture I co-founded changed its mind three times about who the product was for. First it was a score for consumers. Then it was a score for businesses. Then it was something stranger and more useful: a segment-agnostic learning instrument. The third framing is the one that unblocked the roadmap, and I almost talked us out of it for being too abstract.

The reason it took three tries is that we kept asking the wrong question. We kept asking "who is the customer?" when the question that mattered was "where do we learn the fastest?" Those sound like the same question. They are not. One optimizes for a logo. The other optimizes for the truth about what the product needs to predict.

A wedge is a channel, not a customer

The standard advice is to narrow. Pick a beachhead. Find the smallest viable segment and own it before you expand. That advice is not wrong, but it is incomplete in a way that costs you months. "Narrow the customer" optimizes for focus. It does not optimize for learning. You can pick a perfectly narrow segment and still build a product blind, because narrowness tells you who to sell to, not what real usage will teach you.

Here is the reframe I now use. A wedge is the delivery channel that lets you learn the most about the problem from real usage, and then expand from. The customer is downstream of that. When we stopped asking "consumer or business?" and started asking "which channel puts our output in front of a real decision, fast?", the answer stopped being a market and became a mechanism. We did not need to decide who the score was for in some final sense. We needed to find the one place it would get used, because usage is the only thing that tells you whether the thing you built predicts what you think it predicts.

This is why "segment-agnostic learning instrument" unblocked us. It freed the roadmap from a premature commitment. We no longer had to win a vertical to make progress. We had to find any channel where the output touched a live decision, and let that channel teach us. The segment could come later. The instrument had to learn now.

The proof point is a changed decision

So what is the actual near-term goal? Not a closed deal. Not a signed logo. Not a marquee partnership you can put on a slide. The goal is narrower and harder: one real customer used the thing inside their workflow, and it changed a decision they make.

I want to be precise about that bar, because it is easy to clear the watered-down version. A demo is not usage. A pilot that sits in a tab is not usage. A customer saying "this is interesting" is not a changed decision. The bar is that someone with a real choice to make - to approve or decline, to price high or low, to act today or wait - made that choice differently because of your output. That is the proof point. Everything short of it is theater.

Why is that worth more than a logo? Because a changed decision is the only signal that tells you what the product actually has to be good at. When a real decision moves on your output, you learn three things at once: what the customer was actually deciding, which part of your output they leaned on, and where it was wrong enough to scare them. A signed partnership tells you none of that. It tells you someone in procurement liked the deck. You can win the logo and still not know what to build.

Learning beats channel-lock

There is a trap on the other side of this, and it is the one ambitious teams fall into. Once a channel starts working, you want to wire yourself to it. Build the deep integration. Sign the exclusive. Become the platform. Channel-lock feels like progress because it looks like commitment. It is the enemy of the thing that got you here.

The whole point of the wedge was learning velocity. The moment you optimize for channel depth over learning breadth, you stop being an instrument and start being a feature of someone else's roadmap. You inherit their pace, their priorities, their ceiling. The teams that build platforms too early are usually teams that found one channel that worked and mistook it for the market. They stopped learning and started defending.

So the sequence I hold to now is plain. Find the channel that teaches you fastest. Get one decision-changing use - real workflow, real choice, real movement. Then scale the proof, not the integration. Scaling the proof means finding the next channel where the same instrument changes the next kind of decision, which is how you discover whether you have a product or a point solution. The platform, if it ever comes, is the thing you earn after the instrument has learned enough to deserve one.

That is the discipline behind "wedge first, platform never." Not that you never build a platform. That you never lead with one. You lead with the narrowest thing that teaches you the most, and you refuse to confuse the channel that taught you with the business you are building. The wedge is a teacher. Promote it to anything more permanent before it has finished teaching, and you will scale the wrong product with great conviction.

Field Notes from the Agentic Operator is a personal series. These are my own views, not those of my employer or any organization I work with, and nothing here relies on non-public information.