Investigating qualitative depth of AI-moderated interviews
Summary
Faced with the need to understand our customers’ decision-making (those who purchased our platform, and those who didn’t), I initially relied on surveys. However, I found they lacked the nuance that real conversations reveal. By introducing AI moderation, I created a way for participants to engage in adaptive, conversational interviews that went beyond the limits of static forms.
I’ll walk through how I set up these sessions, what prompts worked (and didn’t), and how I analyzed the results. I’ll also share how I’ve used other AI tools like ChatGPT and Perplexity to assist with synthesis and bias-checking, creating a workflow that both expands my analytical reach and strengthens the rigor of my findings.
As AI tools continue to enter the researcher’s toolkit, this case study illustrates how we can thoughtfully integrate them to expand—not erode—the human depth of qualitative work. It offers a model for how lean teams can maintain research quality while navigating the realities of limited time, budget, and capacity.
This talk will explore the emerging space between automation and augmentation, finding opportunity for depth when time and resources are tight.”
Key Insights
-
•
AI moderation enables more qualitative depth than surveys due to participants speaking freely rather than typing.
-
•
AI moderation platforms require detailed and explicit research plans to guide effective questioning.
-
•
AI moderation often struggles with natural conversational flow and can produce awkward pauses or interruptions.
-
•
Follow-up questions from AI moderators can be vague, repetitive, or miss opportunities to dive deeper.
-
•
AI moderation is best suited for validation or deepening understanding of known problem spaces, not foundational discovery.
-
•
Participants may prefer AI moderation over surveys due to ease and flexibility, supporting higher and quicker recruitment.
-
•
AI moderation platforms currently do not replace human analysis or provide strong quantitative outputs.
-
•
High emotion or sensitive topics may work better with AI moderation if participants prefer anonymity, but human interviews excel at rapport.
-
•
Researchers need to invest time troubleshooting and iterating prompts in AI moderated studies.
-
•
AI moderated studies allow teams of one to scale qualitative research when live interviews are infeasible.
Notable Quotes
"Having one-on-calls with people is one of my favorite parts of conducting research, and it felt like this was being taken away from me just for the sake of saving business resources."
"AI moderation, in short, is a large language model that runs a moderated interview session without a human researcher present."
"I was interested in thinking of this method as a survey plus that you can use to get survey-like data but with more qualitative depth."
"When I gave the model my list of interview questions, it actually made for a really bizarre participant experience because the model just went down the line of the different questions."
"People were more likely to explain their thought process and just more context around the particular situation of their org."
"There were moments like long kind of awkward pauses where the model was processing, and sometimes it actually spoke over participants."
"Follow-up questions missed opportunities to probe deeper or were confusing, and participants had to repeat themselves."
"If I really just needed only quantitative data from a survey, then I would not go the AI moderated route."
"AI moderation did not make the research easier. It made it possible under real world constraints."
"From setup to analysis, AI moderation still requires a good understanding of what quality research looks like and how to interpret qualitative data."
Or choose a question:
More Videos
"What is your mission for this design Ops summit? When this conference ends, what do you want to feel?"
Louis Rosenfeld Bria AlexanderDay 1 Welcome
September 23, 2024
"ChatGPT’s approach in analogy generation segmented the narrative into unrelated metaphors, which confused the overall story."
Weidan LiQualitative synthesis with ChatGPT: Better or worse than human intelligence?
June 4, 2024
"Impact can be about changing the way people talk about the problem and see the customer."
Megan Blocker Mujtaba Hameed Victor UdoewaPanel: Excellence in Impact
March 25, 2024
"Half the battle is having good questions and the right language to ask those questions."
Louis RosenfeldGenAI for UXers: A Rosenbot Demo and Discussion
June 11, 2025
"There will be session notes and sketch notes so you don’t have to take notes yourself unless you want to."
Bria Alexander Louis RosenfeldWelcome
January 8, 2024
"By limiting our classes to six or seven students, we could give better quality feedback."
Marjorie Stainback Kelsey KingmanTransforming Strategic Research Capacity through Democratization
October 24, 2019
"A key metric for a successful design system is how early the design system team is invited into project conversations."
Dan Mall“Ask Me Anything” with Dan Mall, Author of Upcoming Rosenfeld Title, Design that Scales
October 2, 2023
"After 90 days, we had double-digit improvement on our NPS and a decreased operations cost per load."
Alana Washington(Remote) Service Design: A Transformation Case Study
June 8, 2022
"Writing my book was just another UX research project synthesizing my career and leading me toward law."
Chelsey GlassonExit Interview #3: Same as It Ever Was: What Leaving Tech Taught Me About Change
December 17, 2025
