Marketing surveys and focus groups are a notorious grind. You’ve got to find people (and lots of them) in various demographics, persuade them to participate, and pray for insights. Researchers at Columbia, Berkeley, and Alberta universities wondered: Could generative AI replace this process, helping you to understand your customer without ever interviewing a single person?
Related: 8 Ways to Use AI to Keep Your Customers Happy
The research
Could ChatGPT give the same answers that humans give? The researchers tested it.
Because ChatGPT is trained on the internet, its default is to represent the population at large: old, young, different incomes, etc. So to get the AI to answer like all those people, they gave it this prompt: “The most sporty car brand is…”
Next, the researchers asked hundreds of actual humans to finish the same sentence.
In another part of the study, they categorized hundreds of humans by age, gender, and income, and asked them what their favorite car brand is. Then they asked ChatGPT to think like those humans, and answer the same question. For example, a low-income, 30-year-old woman shared her favorite car brand. Then the researchers gave ChatGPT a prompt like “A young woman and poor female’s favorite car brand is…”
The results
The AI pretty much nailed it: The human and ChatGPT answers were the same more than 75% of the time. The researchers went on to compare the answers to car trade-in data — and found that they, too, matched up to 75% of the time.
Related: Will Artificial Intelligence Replace Human Interaction? 4 Ways AI is Impacting and Empowering Customer Experience
What we’ve learned
The results are impressive, though narrow. This test assessed how different demographics view an established product category — but that’s a far cry from, say, how consumers might see a completely new product idea, or what your customers might feel about your product.
How to use this
Experiment! To get AI to respond like your target consumer, start by feeding a tool like ChatGPT all the information you know about a customer segment — demographics, typical social media posts, any messages they’ve sent your company, and so on. Then you can grill the AI doppelganger all you want about their goals, wants, and needs. You can also try more specialized tools, like Crowdwave.ai, which were custom-built for this kind of research.
AI-driven consumer research can be especially helpful for hard-to-reach consumers, says Stefano Puntoni, codirector of AI at Wharton and a professor of marketing. For example, imagine you’ve got a software platform that’s built for procurement directors at big companies. “These people are busy and get hassled all the time — and you can’t just ask any random person on the street,” Puntoni says. “But you can probably create a synthetic profile that’s a pretty good surrogate.”
Puntoni says there are many limitations to AI-driven consumer research, though — for example, in trying to optimize pricing. “We’re still figuring out what this is good for and tweaking things,” he says. “But certainly today, if you’re an entrepreneur trying to understand a particular target market, creating those synthetic profiles and then interviewing them to gain knowledge about these people is already quite valuable.”
Related: AI Might Know What You Are Feeling Before You Even Do — Here’s How AI Can Help Us With Client Feedback
Marketing surveys and focus groups are a notorious grind. You’ve got to find people (and lots of them) in various demographics, persuade them to participate, and pray for insights. Researchers at Columbia, Berkeley, and Alberta universities wondered: Could generative AI replace this process, helping you to understand your customer without ever interviewing a single person?
Related: 8 Ways to Use AI to Keep Your Customers Happy
The research
The rest of this article is locked.
Join Entrepreneur+ today for access.
