By: Daniel Graff-Radford, CEO, Discuss
Jo Lindenberg, Director of Global Consumer Insights at HelloFresh, describes herself as an AI skeptic. In her words, “I think there still is a lot of faff around it.” Then results from their UK pilot of AI moderation were stronger than expected. Jo calls this the “stranger on the bus” phenomenon. Without a human moderator to perform for, respondents simply open up. “There’s not a person that I have to prove something to,” she said.
Jo’s team works with Discuss, which is why I can tell you what the internal debate looked like before the pilot ran instead of just the outcome afterward. Her finding is not an outlier. A market research agency specializing in public health and healthcare access has found the same pattern: on questions involving HIV, suicide prevention, and sexual health, respondents say more to an AI interviewer. “They will say things to an AI interviewer that they would never say over the phone.”
Two independent researchers, different industries, the same counterintuitive finding. In some contexts, AI creates a quality of disclosure that human-led research cannot. It was one of the most important takeaways from a room of 40 research leaders in New York this past April, and I want to tell you what else that room taught me.
I Walked in with a Hypothesis. The Room Revised It.
This past April, Discuss hosted a customer event in New York. The group spanned healthcare, consumer goods, financial services, and media. As CEO of Discuss, I went in thinking AI interviewing was primarily an efficiency story: faster exploration, lower cost per respondent.
At the beginning of the session I asked the room a question: how many of you would currently use AI interviewing for a real project? Fewer than 10 percent of hands went up.
By the end of the session, almost every person in the room wanted to test it.
That shift happened because the conversation surfaced something the simple AI-versus-human debate tends to obscure: what each approach does best, and where in the research process each one creates the most value.
The Fear Underneath the Hesitation
Why was that first show of hands so low?
Insights professionals who have built careers on the craft of moderated IDIs are right to be skeptical of any cheaper, faster substitute. If AI interviewing is framed as a replacement for skilled human moderation, it erases what the researcher does rather than extending it.
A leader at a major healthcare organization described her situation directly: by the time her team delivers findings through the traditional process, the business question has already moved on. Her insight delivery cycle runs nearly 10 months. Her stakeholders are making product decisions in weeks. She needs her researchers focused on the questions that genuinely require their expertise, with something else handling the exploratory, early-stage work.
When the room heard that framing, the defensiveness gave way to curiosity about where AI interviewing actually fits.
Where Does AI Interviewing Fit in the Research Process?
Here is the model that emerged from the roundtable, articulated more clearly by the participants than I could have articulated it going in.
AI interviewing belongs between quantitative surveys and live qualitative interviews. Quant tells you what is happening at scale. Live qual, run by a skilled moderator, tells you why it’s happening with a depth and interpretive richness that it was built to deliver. AI interviewing gives you directional signal, early-stage validation, concept screening: fast and broad, good enough to determine whether a question is worth the investment of a deeper study.
The practical implication is that AI interviewing produces the work that makes the skilled moderator’s time worth spending. If an AI interview surfaces five themes across two hundred respondents, the expert researcher knows which two of those themes warrant a live conversation, and with whom. Breadth handled by the tool; depth handled by the person who knows how to ask the follow-up question that actually matters.
Keeping the same participants across both modes changes what the live session can do. When you recruit people who responded in an AI session into a live moderated follow-up, the moderator walks into the room already knowing what the respondent shared previously.
Can AI Interviewing Reach Places Human Moderation Cannot?
Returning to Jo and the agency working in public health, in some research contexts human presence is the barrier. AI interviews grant a different category of access than what speed or cost improvements can deliver. The traditional method couldn’t produce it regardless of how much time or budget was allocated.
This deserves serious attention from methodologists. The absence of a human changes the conditions of disclosure, which means validation frameworks built for traditional moderation may need rethinking. What the finding ruled out for me is the simple framing on both sides of the AI debate. In some use cases, AI interviews can accomplish what human moderation cannot.
When Should Insights Teams Use AI Interviewing Instead of Live Moderation?
The researchers who raised their hands at the end of the session were persuaded by the possibility that AI interviewing could handle a category of research work their teams currently have no good answer for: the fast, directional question that doesn’t justify a full fielding cycle but needs something more rigorous than a guess. That’s a narrower and more honest claim than the broad efficiency argument, and in my experience it’s the one that holds up when a skeptical researcher actually tests it.
What I hadn’t anticipated was the fidelity case. In the right contexts, AI interviewing doesn’t just get you an answer faster. It gets you a truer one.
The skilled moderator is not in competition with the AI interviewer. They are serving different moments in the same process. The organizations that understand this stop asking which one to use and start asking which moment they’re in.
Daniel Graff-Radford is the CEO of Discuss, an industry leader that unifies AI with qualitative depth and quantitative scale. Under his leadership, Discuss was named a Forrester Wave™ 2026 Leader in Experience Research Platforms. A named inventor on multiple technology patents, Daniel also serves as an Executive in Residence for the Mayo Clinic’s AI initiatives.
