Like many marketing and insight professionals, I have spent the past year exploring ways to do less work and have more fun. LLMs like Claude and ChatGPT, and especially our own proprietary LLM strat7GPT, have proven to be exceptionally convenient for sorting and analysing quantitative data. While valuable and rewarding, these tasks are not exactly what get me up in the morning.
I have chucked many questions into the AI void to see what gets spat back out, for example: “How do I rebase this against total sample?”, “To what extent is the sentiment towards the campaign positive from reading these open ends?”, and my favourite “Rewrite this sentence to make me sound smarter.”
However, while the time-saving benefits of AI for quant data are obvious, I was initially dubious about its potential for qualitative research. Outside of general project set up, I felt that AI didn’t belong at the qual party. Surely qual is all about human interaction, with the researcher coming away from an interview, focus group or assisted shopping trip with a deep human understanding of each consumer’s wants and needs? So, when the opportunity arose to test a new AI moderation tool as a proof of concept, I was more than intrigued.
The tool in question was Tellet, an AI interviewing platform designed to ask qualitative questions via text message while encouraging respondents to reply via voice notes. Their AI moderator is trained to ask probing questions as necessary, depending on the overall objectives of the project and the participants’ response.
After a demo and rigorous testing, we set our newly named AI bot, Gary, upon consenting participants. We tasked Gary with finding out the barriers to usage for a well-known loyalty scheme. And off he went.
As we had briefed our participants that they would be conversing with AI via screening calls, the initial responses rolled in pretty quickly. Many were eager to compare the experience of an AI-moderated interview to a human one.
Feedback was generally positive: participants appreciated the ability to respond at their own pace, on their device, at a time that suited them. Many were also complimentary of Gary’s listening and comprehension skills – something they were not anticipating to work so well.
However, we did notice that the novelty of speaking to an AI bot wore off for some, and we had to chase a few final completes. This was a valuable learning experience, proving that, unsurprisingly, people are just not enthusiastic about certain topics.
Subjects like finance and loyalty schemes often require a human moderator to “jazz-hands” their way into an engaging and enlightening interview. But for most of our participants who were happy to tell Gary all about their experiences, the interviews were a success. And for us insight professionals, the fun was only beginning.
Tellet not only facilitates the interviews – it also has a powerful dashboard and analytics toolkit for us to pore over as the responses funnelled through. This meant we could analyse sentiment, identify common themes, and spot outliers within the transcripts.
Much like uploading an interview or focus group transcript to an LLM to kickstart thematic analysis, Tellet highlighted key observations as they happened, keeping us informed with a smart dashboard. This made it easy and quick to analyse the data in situ, forming a solid starting point for our analysis.
As this was a proof of concept, we spent a lot of time verifying that Gary’s understanding of the transcripts was correct. Except for one or two instances, Gary really seemed to “know” what he was talking about. Better yet, he also made suggestions as to what this meant for our client based on the objectives we had trained him to focus on over the course of the research.
At this point, I started to get a little worried. Who did Gary think he was coming here and making my job look so damned easy?
Then I had a cup of tea and calmed down.
The role of an insight professional is much broader than simply gathering insights quickly for clients and putting them into a pretty dashboard. A massive part of our role is understanding when and why to use research tools from the growing arsenal of tech at our disposal.
An AI moderator is not going to cut it for more complex or sensitive projects, for example, among audiences like the elderly or those without personal devices. AI focus group moderation also remains tricky – it lacks the skills a human moderator needs to navigate chatty group and balance conflicting opinions.
Us insight professionals also must interpret results in ways that not only make sense on paper, but make sense in practice. Gary isn’t going to know everything – he only knew what we told him one afternoon while plugging in the project objectives. So, while his AI-generated recommendations helped guide our thinking, they lacked understanding of the broader loyalty scheme experience – elements that were discussed between us humans over the course of the project.
So, if you’re an insight professional worried about AI taking your job, don’t be.
Tools like Tellet simply free us up to do the fun stuff and less of the boring stuff. AI qual tools aren’t going to replace human moderators, who will always be needed to provide broad, sensible industry context, deeper human understanding, and, most importantly, to wave their jazz-hands during an interview about insurance.
Sarah Whelan
Insight Manager
Reading Time: < 1 minutes
Reading Time: < 1 minutes
