Product Discovery and Customer Validation: New Version of the Skill in Skill Arena

Why a new version

In June 2026 we published the first version of this skill, focused on the classic core of the discipline: interviewing without bias, synthesizing, mapping assumptions, experimenting, and deciding. That core remains valid. What has changed in just a few months is the terrain on which it is practiced.

Artificial intelligence is already part of the daily work of discovery. According to Maze's Future of User Research Report 2026, published in March, 69% of those who research with users use AI in at least part of their studies. It transcribes, summarizes, groups themes and, on some platforms, conducts the interview without a human interviewer. At the same time, prototyping tools make it possible to build in an afternoon what used to take weeks. Marty Cagan summarized it in July, citing Hilary Gridley: since it has never been so fast to build, it has never been so easy to run ten times faster in the wrong direction.

When building is cheap, the bottleneck shifts to knowing what to build. That is the underlying reason for this version: product discovery does not lose value with AI, it gains it, and it demands new criteria so that speed is not confused with learning.

What's new

The skill is now organized into six blocks and eighteen topics, and each topic has its own chapter in the development guide.

The four blocks of the cycle are kept and expanded:

• Converse and capture signal, with three new topics: separating the problem space from the solution space, deciding whom to listen to (segment, sample, and recruitment), and consent and data handling when recording and transcribing interviews.
• From data to understanding, with synthesis that moves from observation to insight, and framing the problem as the job the customer wants to get done.
• Design experiments to learn, with assumption mapping, prototype tests, and the MVP understood as an experiment, now also when AI makes building cheaper.
• Decide with evidence, with success criteria set before launching, strength of evidence as a gradient, and the decision between persevering, iterating, pivoting, or stopping.

And two blocks are added:

• Discovery as a team habit: the weekly cadence of the product trio, continuous recruitment, the evidence repository, and the opportunity solution tree that connects what has been learned with the outcome being pursued. The practice is far from widespread, because in ProductPlan's State of Product Management 2026 survey only 34% of teams collect customer knowledge regularly and use it to prioritize.
• Discovering with artificial intelligence: where AI helps when interviewing and synthesizing and where it erodes, what place synthetic users occupy on the evidence scale, and how to validate a product whose core is a model.

Three ideas the new version puts front and center

Compare, don't delegate. The most widespread antipattern at the moment is dumping transcripts into an assistant and accepting its summary. The guide proposes the workflow Teresa Torres advocates: synthesize first, then ask the AI for the same synthesis and contrast the two before asserting anything. A tool's summary is one more interpretation, and it is treated as such.

Synthetic is not evidence. Synthetic users are useful for preparing the real work, such as desk research, hypotheses, or rehearsing a script, but their answers are opinion without cost. The academic evidence from 2026 is negative: a study published on arXiv in July found that models inflated differences between segments by two to four times and led to choosing the wrong segment in half of the cases studied in the United States. On the strength-of-evidence scale, synthetic sits on the lowest rung.

Recording also carries obligations. Recording, transcribing, and analyzing an interview means processing personal data. The Spanish Data Protection Agency published its criteria on AI voice transcription in January and April 2026, and since August 2 the European AI Regulation requires that a system interacting with people inform them that they are dealing with an AI. The guide turns this into a simple practice that fits in the script of any interview.

What we have corrected

A new version is also an opportunity to correct. When reviewing the first version against its own sources, we found three ideas that had settled in without their origin, and we have corrected them:

• The template "when [situation], I want [motivation], so that [outcome]", often presented as the format of jobs to be done, is the job story, born at Intercom and named by Alan Klement. The jobs-to-be-done framework also has two schools worth distinguishing: that of Clayton Christensen and Bob Moesta, and that of Tony Ulwick.
• The assumptions to test are not just three. To desirability, viability, and feasibility, Teresa Torres adds usability and ethics, and David Bland adds adaptability. They are prioritized in their importance and evidence matrix.
• Faced with evidence, the options are persevere, iterate, pivot, or stop, and changing segment is one of the types of pivot described by Eric Ries.

Who it is for and what it certifies

The skill complements Scrum Manager's Product Owner training, which teaches what continuous discovery is and what place it occupies in the product. Here the how is trained: the protocols, thresholds, and decision criteria. It is useful for product managers, designers, developers, Scrum Masters, and analysts who take part in discovery.

The Skill Arena diploma certifies the ability to obtain reliable signal from customers, test the riskiest assumptions with the minimum experiment, and decide with the evidence at hand while declaring the evidence that is missing. The training and assessment banks have been fully regenerated from the new guide, at two difficulty levels.

The new version is available in Spanish and in English.

How to access it

The State of the Art, the development guide, and the training and assessment platform are on the skill's page in Skill Arena: Product Discovery and Customer Validation.

Read, train, and tell us what you think. Every contribution, criticism, or suggestion helps us prepare the next version.