Numbers show where visitors leave. They do not say why. The usual answer is a round of customer interviews, which can take weeks to recruit, run and write up, so it happens less often than it should. Henkan Research is an AI user researcher built to make that round routine. This post describes what Research does today, in beta, without the roadmap.
Prepare a study with a coach
A study answers one business question. You start by describing the problem and the decision the study should inform, then write the topics to cover. A coach, built to work like a senior qualitative researcher, sits next to every preparation step: it explains what the step is for, suggests two or three improvements (a sharper decision, a new hypothesis, a better opening question) and answers your questions in plain language. You do not need research training to get a usable guide.
Interviews last 10, 15 or 20 minutes, with a default target of six completed interviews per customer profile. You can add visuals, such as a page or a mockup, to show at a given moment. Before inviting anyone, you can take the interview yourself and judge the interviewer first-hand.
Interviews that follow up
Participants can be invited by email, through a public link or QR code, or by a small invitation shown on your website, on the pages and at the moment you choose. Each participant talks with the AI researcher whenever it suits them, out loud or in writing, in French or English, and, when the study offers it, can switch from voice to writing at any time.
The interviewer introduces itself as an AI, then runs a semi-structured interview: your topics are the plan, the conversation stays natural. It starts from lived experience ("tell me about the last time…"), asks one short, open question at a time, and digs deeper on a rich point with up to three follow-ups. It shows your visuals when the topic comes up and waits for a spontaneous reaction. It manages time, covers every required topic before closing, and sums up what it understood before thanking the participant.
There are also things it never does: ask leading or double-barrelled questions, state your hypotheses, defend the brand, or ask for sensitive data such as health, finances or passwords.
From transcripts to a synthesis you can check
Every interview comes back as a transcript with topic coverage and a timeline. Ask your interviews answers questions across all of them ("who mentioned delivery costs, and when?") and each claim cites the participant and the moment, one click away.
The synthesis turns completed interviews into findings for each objective, a verdict on each hypothesis (confirmed, refuted or unclear), ideas participants raised on their own, reactions to visuals and two to four A/B test ideas. Every statement is backed by quotes linked to the exact moment in the interview. With few interviews, it gives counts such as "3 of 5 participants" rather than percentages.
The synthesis starts as a draft. A person reads it, corrects it and validates it; only then can it be shared with a read-only report link.
Where Research stands
Research is in beta and available to Henkan consultants and the teams of Henkan's clients. Every new organisation starts with a demo study, so you can see every screen before running your own. The AI interviews and summarises; your consultant frames the question and turns the findings into tests.
If you already have access, the documentation follows a study from the first brief to the shared report. If not, talk to Henkan.
