How I Tested That · October 14, 2026

The Build Trap With AI: Melissa Perri on Deciding What to Build First

The build trap is when a company keeps building features without asking if they are the right ones, and never checks whether they worked. Melissa Perri describes the build trap in her book Escaping the Build Trap. In an October 2026 interview, Perri says AI lets companies fall into the build trap faster, because building software is now cheap. To escape the build trap, Perri says to set the strategy before you build. Then keep people who can judge the AI's work, pair each cost goal with a guardrail, and talk to users before you trust the data. She said this on David J. Bland's podcast How I Tested That, published 14 October 2026.

Stuck on a real call of your own? Gavel answers startup questions with what a named founder did, and links to the minute they said it. Ask Gavel free

Speaker
Melissa Perri
Where
How I Tested That
Published
Length
48 min

Melissa Perri: Author of Escaping the Build Trap, founder of Product Institute. Interviewed by David J. Bland.

Recording: Melissa Perri | How I Test for the AI Build Trap

How to escape the build trap

  1. 1

    Decide what to build before you build faster.

    Perri gave her first talk on the build trap in 2015, when the promise was to build faster with Agile. She says AI is the new shiny tool. Big companies now spend hundreds of millions of dollars on AI and expect it to multiply their output. The money is there, she says, but nobody has worked out what to build with it.

    “they're not spending the time and the dedication actually figure out what we should be building with the AI.”
    Melissa Perri, How I Tested That, 04:17
  2. 2

    Keep the people who can judge the work.

    AI can write a spec fast, Perri says, but someone has to read it and know if it is any good. When her team rebuilt its website with Claude, the specs came back with things nobody asked for. She tells of a large bank that laid off its mid-level and senior product managers. It kept the cheaper product owners to steer the AI. She says most product owners she works with can't set the strategy for a large product.

    “we still need to steer the AI. And to steer the AI, you need to have people who have good judgment,”
    Melissa Perri, How I Tested That, 10:00
  3. 3

    Pair every cost goal with a guardrail.

    Perri says a goal such as cutting customer service costs is so broad that almost any project fits under it. So she pairs it with a measure that must not drop, such as customer satisfaction, NPS or retention. She calls these mutually destructive pairs. She says companies forget the guardrail when they ask AI to cut costs.

    “I talk about mutually destructive pairs”
    Melissa Perri, How I Tested That, 22:30
  4. 4

    Wire up your data so you can see what worked.

    Perri says many leaders can't tell whether a project they paid for is moving the goals set for it. Once the data is clean and the tools are linked, anyone in the company can ask it questions without writing code. Early in her career she took a seven-week course just to pull data out of her company's database. She says product ops should own that setup.

    “anybody in your companies can start asking those questions”
    Melissa Perri, How I Tested That, 17:47
  5. 5

    Talk to users before you trust the data.

    One company Perri worked with had usage data that said therapists did not want its product. When the team asked them, the therapists said they did want it. It just broke so often that pen and paper was faster. Perri uses AI to size markets and sort research notes. But she says AI will never set your strategy, because it can't learn why people act as they do.

    “If you only looked at the data, it would just tell you nobody wanted to use it”
    Melissa Perri, How I Tested That, 33:03
  6. 6

    Set the strategy before you build an AI team.

    Perri has seen seven companies set up a separate AI and data department, often under a chief AI officer. Then they look for problems to give it. She says to work the other way round. Start with the strategy and find where AI could help your customers in a new way. Then hire the team to build it.

    “rather than let's start from the strategy,”
    Melissa Perri, How I Tested That, 43:47

Where experts disagree

Paul Graham says how fast a startup ships is still the best sign of whether it will succeed, and that AI has not changed this. Perri builds her own small apps with AI and likes the speed. Her worry is big companies that build fast with no clear idea of what to build.

Paul Graham: shipping pace still predicts success

Eric Ries says customer talks give you ideas. To him, only measured behavior can prove an idea. Perri's therapist case cuts the other way. The usage numbers said to shut the product down. Only the interviews showed why people had stopped using it.

Eric Ries: validate with measured behavior

Which call fits your case

  • If you are a founder who already knows where your product is going, Perri says AI tools let you build it quickly.
  • If your company wants an AI team before it has a strategy, write the strategy first. Hire the team after.
  • If your goal is to cut a cost, pick a second number that must not drop, like retention.
  • If your usage data says nobody wants a product, talk to the users before you shut it down.
  • If your strategy is set and the open question is pace, follow Paul Graham and ship faster.

How it compares

How this recording lines up with what Melissa said before and with other operators Gavel cites.

Agrees with

Jake Knapp makes the same point about building with AI. He writes a clear plan of what the product needs to look like. Only then does he prompt an AI tool, so the result does not come out generic.

Jake Knapp: plan before you prototype with AI

Builds on

Perri's Strategy Alignment Test asks every team what it is building and why. In this recording she asks the same of AI budgets. When a company says it will spend a set amount on AI, she asks what the money is for. She says the answers are often confused.

Melissa Perri's Strategy Alignment Test

Related

Stuck on a real call of your own? Gavel answers startup questions with what a named founder did, and links to the minute they said it. Ask Gavel free

New experts every week

Know when the next expert lands.

Gavel adds new operators to the database every week, each one with cited frameworks you can check and a note on where they disagree with the others. You found this page by searching. Get the next one by email instead.

Latest: Ethan Smith on Answer Engine Optimization

One email a week, only when new experts shipped. Unsubscribe with one click. We never sell or share email.