AI-moderated interviews behind-the-scenes!
If AI-moderated interviews make you feel overwhelmed or a bit scared, I totally get it. I’ve felt that same fear while figuring out how to use them well.
The way I’ve started to think about it goes way back to when my stakeholders asked me to use surveys for literally everything. I pushed back every time. When they asked for jobs to be done, I pushed back again, and when unmoderated usability testing came along, I was scared of that too, because I didn’t understand how to do it well yet.
Every single one of those ended up in my toolkit once I learned how to use them with intention, and I think AI-moderated interviews are the same. What excites me most is being able to hear from a lot more people than I could ever interview myself, in their own words, with follow-up questions a survey could never ask.
I recorded a behind-the-scenes video in partnership with Typeform, where I build a study in Research Flow from scratch and then go through the results of the study I ran on how researchers are using AI and what they’re keeping for themselves (I know, very meta!).
The TLDR if you’re short on time:
I start every AI-moderated study with a goal and the decision it’s meant to inform
I keep sessions to 10 to 15 minutes, with five to seven questions on one narrow topic
You can follow up on a multiple choice question, which is something I’ve dreamed of forever in surveys
I always write follow-up guidance. It can feel like I’m over-indexing on instructions, but I would much rather over-index than leave the moderator guessing
The results give me breadth across every session and a highlight reel (clipping clips is the death of me as a researcher), plus topics I can trace back to what people said, so I know where to dig deeper
I see AI-moderated interviews as a starting point that I triangulate with 1x1 interviews, usability tests and concept testing
I always run out of time on these (I could talk about this stuff forever), so I brain dumped everything else into a (free) downloadable guide. It covers how I decide whether a study is a good fit for an AI moderator, the models and fill-in-the-blanks I use, and a lot of examples: