Questions About Using Generative AI for Voice of the Customer Market Research? We Have Answers!

By Kristyn Corrigan, Principal and Practice Lead at Applied Marketing Science (AMS) 

Curious about using Generative AI for market research and product development? I recently led a webinar facilitated by the Product Development and Management Association (PDMA) titled, “Breaking Boundaries: Harnessing AI to Fuel Innovation.” 

I shared some exciting developments from AMS research currently in progress in partnership with Dr. Artem Timoshenko of the Northwestern University Kellogg School of Management.  

Webinar attendees asked some fascinating questions – read on to learn more. 

What is the best type of text data for Generative AI to collect customer insights from? 

When we’re thinking about generative AI for VOC research, it's important to remember that we’re looking for depth and trying to get beyond the obvious with customer needs. Therefore, text data where customers are telling stories and sharing their experiences, emotions, and needs in great detail works best. For that reason, the best text data to use for VOC with generative AI would be traditional interview transcripts or focus group transcripts.  

Customer reviews can also be helpful because a customer typically leaves a review when they’ve had a very positive or a very negative experience. Less helpful would be text that is not expansive or descriptive – such as social media posts. For example, if you were to go look at the comment section of an Instagram post for customer verbatims, you’d likely see comments along the lines of “So great,” “Cool,” or “Love it.” While that expresses some sentiment, it does not express a customer need. Text with rich detail, stories, and experiences tends to work best. 

How can Generative AI be used for scanning social media? 

There are a lot of tools out there right now that will scan social media for you. Some of the sources that I’ve found to be most helpful are specific customer forums or special interest groups. I’ve found interesting discussions and insights on Reddit – there are subreddits for almost anything you’d need data on, from business categories to patient groups. The key thing to keep in mind is that the data you have contains enough detail and depth about customer experiences and pain points. 

Can AI be used to conduct customer interviews, or augment the interviewer in asking contextually appropriate questions? Could AI be used in the customer discovery process? 

There are some off-the-shelf tools that have AI-enabled features, but in my personal opinion as a VOC consultant, there are certain aspects of VOC that are a value-add. One of those is the act of sitting down with your customers and either interviewing them or visiting them – being a part of that experience by listening and/or being present with them.  

AI can be helpful for high-level customer interview summaries and transcription, but there is value in hearing first-hand what your customers are asking for. It creates goodwill to visit your customers and hear about their problems and pain points in a non-sales context. AI may augment the customer interview process, but I think with any of these tools you should ask yourself where AI brings the most value and where we, as researchers, should be putting in effort.  

I’m sure there are parts of the discovery process that could be automated, and I can imagine these machine learning tools being particularly helpful in, for example, a quantitative survey where there are some adaptive survey methodologies – but I do think of AI more as a co-pilot.  

There’s value to having that human touch, especially in the interviewing and observation portion of customer discovery, as well as in the synthesis of results. The act of reading interview transcripts or sifting through customer verbatims is time-consuming and I believe AI could be helpful in leveraging our time. 

Does Generative AI replace traditional VOC efforts? 

I view a fine-tuned LLM as an optimizer, rather than a replacement, for traditional VOC. I believe we must keep a pulse on what these AI tools can help us with in our day-to-day efforts, and how they can make us more effective and able to focus on value-add tasks. That said, it is critical to have the correct data input to the model; by that, I mean the customer experiences and verbatims that get at their underlying needs.  

As I mentioned earlier, I believe it is still necessary in many cases to visit and observe your customers to understand those latent or unstated needs that AI will not be able to glean. I don’t think an LLM will completely replace traditional VOC, but I believe it can support VOC and help overcome some barriers to conducting VOC systematically.   

Interested in learning more about the future of AI for VOC? Check out the full webinar to hear how Generative AI can help to accelerate and enhance customer insights – now available to watch on demand: 

Have any industry- or organization-specific questions about AI and customer insights? Get in touch with me via email, or put a few minutes on my calendar to chat.