Every stage below is something someone is already selling. What almost nobody sells is the whole loop, run by one researcher who owns the methodology. That is the opening.
A line, not a loop. The knowledge stops accumulating the day the engagement ends, and the next study starts from zero.
Six stages. The grey label marks where the machine does the work. The cognac label marks where the value is, and where it stays.
The question, the sample, the instrument, and the vocabulary the whole engagement will use. Definitions agreed before anything is measured.
AI-moderated interviews running in parallel at a scale one moderator cannot reach, plus conversational survey instruments. The guide is still authored.
Transcription, coding, tagging, cross-tabs, stat testing. Everything mapped to the vocabulary set in stage one, so studies can talk to each other later.
What matters, what is noise, what it means for this client's decision. The judgment that makes findings defensible enough to act on.
The findings go where the client already works, not into a file. Their team asks questions of the research in Slack or Teams and gets answers grounded in it.
What the client's team asks, and what the research cannot answer, is itself a finding. The gaps become the brief for the next study.
One accumulating record. Every study writes into it, every question reads from it. Study three is worth more than study one because it can see the first two. This is the only piece worth owning permanently, and it is the piece that becomes an asset rather than a series of invoices.
By a wide margin the leading development researchers expect. Already in use for pattern finding, transcription, and writing deliverables.
Organised, accessible research as the infrastructure that lets insight spread beyond the person who commissioned it.
Ongoing, integrated research rather than a sequence of discrete engagements that each start from nothing.
The AI moderation platforms sell to corporate insights departments. The repositories sell to product organisations. The analysis tools sell to whoever already has the data. Each one solves a stage and hands the client the integration problem.
Nobody sells the loop. Assembling it takes methodological judgment about what to ask, a vocabulary consistent enough that studies compound, and someone accountable for whether the answer is right. Those are not features. They are what twenty years of running studies buys.
Which means the same work can be sold twice: the study, and the thing the study leaves behind. The client keeps paying for the second one because it keeps answering questions after everyone has gone home.