We use AI openly, but with strict bounties.

We use it for transcription on most projects, and transcripts are part of what you get. One honest caveat: recordings made in the field are messier than remote interviews, so on contextual work we use a mix of transcripts and researcher notes taken in the moment rather than relying on transcription alone.

We use it to draft discussion guides and screeners, then rewrite them. First drafts come back too broad and produce polite, unhelpful answers. It is a starting point, not an output.

We use it to check our own interpretation, by asking what we might have missed in a set of findings. That catches blind spots.

On UX audits we actually recommend running the AI pass first, to clear the obvious accessibility and best-practice issues quickly, then applying human judgement to what is left.

Where we do not use it is in the room with participants, particularly on sensitive work. A neutral presence is not the same as a safe one, and knowing when to stop asking is not something we would delegate.

We also check how any tool we use handles what we put into it. Your recordings are your customers’ words.