More on Gen AI Usage in Entrepreneurship

For the last 2 years I’ve made gen AI mandatory in the classroom, but this semester there were so many useful AI tools and other techniques that I brought them in more heavily.

A few examples from a class of grad students testing and potentially launching new businesses:

Problem: I had the student founders start off with more practice on customer discovery interviews. Many of them have done interviews like this in the past, but I often noticed that they either never knew how to do them well or forgot in the moment. The problem was partially educational, but mainly in terms of getting the practice of how to do them well. And that’s the kind of thing you don’t have enough time for in a class.

Solution: I have founders train an AI to analyze their interview questions and give feedback, specifically looking for common problems like hypothetical questions, forward-looking statements, and more.

Problem: Founders do their discovery interviews and then don’t adequately record responses. As long as I’ve done these I have relied on handwritten notes, whether paper or laptop. But in the past when I sat in on these interviews (pretending to be part of the team) I noticed many different things than showed up in the notes. Also, how do you generate transcripts without spending hours typing?

Solution: Now we record the interviews. I’ve been convinced that enough people have become comfortable being recorded that it’s doing just that. Afterward, we analyze the recordings. Audio alone is fine. That gives a deeper understanding of what people are revealing along the way. And the transcripts can be ~80-90% done with AI transcription.

Problem: How do you really assess whether you have a potential and ready customer? Many times it seems like the founders will just move ahead on anything. That’s actually ok as it’s their decision how to spend their time, but if they want to move in the direction that feedback is guiding them, they should think about what they’re learning through the process.

Solution: For this I use a technique developed by Mike Vladimer and Adam Berk – looking for “people in pain.” Specifically, they recommend building a quantified pain histogram and charting the positive, negative or non-existent pain uncovered. I’ll also be interested to test whether a trained AI is good enough at evaluating these interviews.

You don’t have to use every new tool, but try some of these out.