Discovery is the distance

Discovery is the distance

Janne Lammi's Product Circle survey asked 309 leaders where AI has the most impact. Engineering scored 50%. Discovery scored 13%. That gap is a measurement of where teams are investing, and it explains what happens next.

When execution took quarters, discovery hid inside it. Research happened between sprints, user interviews between builds. The cost of learning was bundled into the cost of building, and nobody separated them on a line item. Execution was the large number. Discovery was invisible overhead.

Now execution is nearly free. An NBER study of over 100,000 developers found that autonomous coding agents increased commits by 180%, but releases rose only 30% (I covered that asymmetry in a previous post). The build accelerated. The judgment didn't.

Something quieter happened alongside it. Discovery became visible. On many roadmaps it is now the largest remaining human cost. Teams always cut the most visible line item first.

The cuts arrived

Perspective AI found that 71% of organizations now have non-researchers conducting studies. Forty-nine percent of researchers are pessimistic about UXR's future, a 26-point jump from 2024. Google Cloud is reportedly cutting all UX researchers below L6.

The logic sounds reasonable. If you can build a feature in a day, why spend two weeks researching whether to build it? Ship it, measure it, iterate. The prototype is the research.

Jeff Gothelf named what this produces: code that "will 'work as designed.'" He called discovery "a nice-to-have and all too often optional part of the software development process." That was before execution costs dropped further. The pressure to skip discovery only increased.

The distance moved

Distance to first proof asks: how many days until a real person uses a real version and has an opinion? When execution was expensive, most of that distance was build time.

Shrink construction and the distance should collapse. It doesn't. The distance includes finding the right person to test with and knowing what their reaction means. Those are discovery costs. They were always there, hidden behind the larger number.

Cutting discovery doesn't shrink the distance to first proof. It removes the team's ability to measure it. You ship fast, but blind.

Pendo's 2019 analysis of 615 SaaS products found that 80% of features are rarely or never used. Fifty-six percent never used at all. That data predates AI-accelerated development. The ratio gets worse, not better, when teams ship features in hours without asking whether anyone needs them.

The cost surfaces elsewhere

Piyush Puri, writing in Forbes, tracked customer acquisition costs climbing roughly 40% between 2023 and 2025. His diagnosis: "The product itself wasn't the problem. What never got tested was whether the market could actually discover the product."

When building is cheap, teams build more. Each launch enters a market where attention is finite. The features that survive are the ones where someone figured out who needs this and how they will find it. That work is discovery. It migrated from the build budget to the acquisition budget, where it costs more and arrives later.

Janna Bastow of ProdPad: "AI doesn't eliminate bad product decisions. It amplifies them." She identified the failure mode: "A well-written spec looks like a well-researched spec." The output quality is indistinguishable. The input quality is not.

The wrong bet

Lammi's survey found that 87.7% of product teams use AI coding assistants. Only 36.1% say it strengthened how they work. He wrote: "When building gets cheap, choosing gets expensive."

Teams cutting discovery are betting that speed to market replaces understanding of market. That shipping ten features and measuring which stick is cheaper than researching which three to build.

Bastow's counter: "Speed is no longer a differentiator." Every team can build fast now. The advantage belongs to the team that builds the right thing. "The shift from delivery velocity to learning velocity is one of the most important transitions a product organization can make."

Discovery is the distance to first proof. The whole distance, once the build compresses. Removing it from the budget does not remove it from the product. It means the 80% of unused features get built faster than ever, by teams that cannot tell which 80% they are building until the acquisition costs arrive.