Continuous discovery means learning from users every week, not once a quarter. Here is what the weekly loop looks like and which parts an AI teammate can take on.
What continuous discovery means
Continuous product discovery is the habit of learning from users every week instead of in one big research project a quarter. Teresa Torres popularised the idea as small, regular touchpoints with customers, run by the team that builds the product.
The point is not the volume of research. It is that decisions about what to build are made with fresh evidence, by the people who will build it, while the evidence still matters.
Why most teams stop after a month
The thinking part of discovery is short. The legwork is long: finding the right users, booking calls, taking notes, checking what people said against what analytics shows, and writing it up so the rest of the team reads it.
That work lands on a PM or a designer who also has a roadmap, reviews and a release. When the week gets busy, interviews are the first thing to go, and the team drifts back to guessing.
The weekly loop
A workable loop has four steps. Listen: collect what users say in support, sales calls, chat and interviews. Find: connect those signals to behaviour in analytics and rank problems by how many people they block. Decide: the team picks what goes into the cycle. Ship and check: make the change and watch whether the metric moved.
Each step leaves an artifact the next one uses: quotes, a ranked list of blockers, an issue with evidence, a pull request, a before and after number.
What an AI teammate can take on
The steps that are mostly legwork are the ones an AI teammate handles well: picking a cohort from analytics, sending invites, running a scripted interview with follow-up questions, transcribing, tagging themes and posting a brief with links to every source.
It can also draft the issue and, for small interface fixes, open a pull request. What it should not do is decide on its own what the product becomes. In Wirey every change waits for a human approval.
How to tell if it works
Count the time from “a user hit a problem” to “the fix is live”. If continuous discovery works, that number drops from months to days, and the team can point to the evidence behind each change.
A second check: ask anyone on the team why a screen looks the way it does. If they can answer with a date, a quote and a number, discovery has become part of how you work.
#product · Monday 9:00Top blockers this week: card asked before trial, 3 required fields on step 1.212 users · 4 sources
PR #128 · Ask for the card after the trial3 files · +42 −13 · lint, typecheck, tests passedWaiting for review from Marc