Transcription
So, got a question from Dyke. Should AI run or rehearse customer interviews? Run them? No. Rehearse them potentially. Help you around the process. Definitely.
So, the reason it can't really run them, even if the voice was better, it is because there's not a great benefit to someone sitting down and answering these questions for you. They do it because it's a way to be heard and to share a piece of their life and to complain or to share their expertise or to share their problems. The moment they realize there's not a caring human being on the other side of that, they are going to be completely disengaged.
So then the way like the naive solution to this would be well let's pay them for their time. But when you pay people for their time, what happens is you don't find out if they actually care because now they've got a reason to answer the questions, which isn't that the topic actually matters to them. And kind of the first and biggest thing you want to figure out when you're building a company is do my customers care? And so there's a hidden benefit to trying to set up the customer conversations and talking to people, which is that like if nobody cares, then no one talks to you. And there's important data in that as annoying as it is. And that doesn't mean the business isn't going to work, but it means you haven't found the the language to make the customer segment realize what you're talking about or you haven't found the customer segment for who that problem or goal or outcome is top of mind. And those are all pretty important learnings.
And if you pay people through for their time, which you would have to do if you were running this as some kind of like automated, you know, machine-assisted interview process, which is basically closer to surveying than to customer learning, it kind of undermines or skips like one of the biggest things you're trying to learn, and you end up learning a lot of details about a problem that people don't actually care about. So, I would not want to have a machine run my interviews for the same reason that I I I don't like using surveys for customer learning.
It's a little bit different if you're dealing with existing customers and you're collecting like case studies, feature requests, bug reports, stuff like that. In those cases, with existing customers, surveys work a lot better. I would imagine that AI facilitated interviews would work a lot better because people then have a reason to be talking to you. They're paying customers. They have a specific problem or goal or or they appreciate the help you've done and and they're like willing to share that story for one of those reasons. Early stage customer learning with non-customers, I wouldn't try it.
In terms of developing your skills as a founder that is more promising. When I teach workshops and and teach, you know, teams how to run this, we often do mock conversations where I give them a random startup idea. You know, one person is the customer, one person is the interviewer, here's your startup idea, here's what it's about. Try to learn about the customer's approach to the problem, their current solutions without talking about your idea. And it's hard for people, like it's hard for them to direct the conversation onto the correct problems without turning it into an implicit or explicit pitch for their idea even when it's just a silly made-up idea I just gave them two minutes ago. And then I have a third person basically act as the observer, kind of taking notes and marking down where the interviewer or the conversation kind of went off track or made blunders. And then you can talk about it and you can switch roles. And in just a few minutes of practice, it's really helpful for identifying what you're doing wrong and where you're going off track.
And certainly, like if you got the AI to kind of sanity check you in the same way where you said, like, hey, here's what I'm trying to do, like here's what I want you to look out for, you know, and you went back and forth. It's like, where did this go off track? I think that is potentially a useful way to accelerate your learning and get some practice reps in before you go into a live conversation.
I think you could also use it after conversations. So if you are doing your conversations online and you are doing a transcript recording, I do not like the idea of dumping the entire transcript into like a knowledge base which you're going to then extract for lessons learned because it muddles together the biased data, the bad data with the good data. I do like transcripts though. It's really helpful to have a full transcript that that you can review yourself and be like, "Oh yeah, this was the exact quote." And like, "Oh, they said something about this. Let me like replay it mentally in my head." So, I like to combine human transcript like review and then decide which parts I'm going to kind of scratch from the record versus save, but really handy. And then I also take my own notes during the conversation itself to kind of highlight the most important things that I want to make sure that I remember to dig out of the transcript or that, you know, are open questions that I want to loop back to later in the conversation itself, that sort of thing.
But if you got the transcripts anyway, then you can also feed the transcripts into an agent afterwards and be like, "Hey, like don't tell me what this means for my business, but I want you to be my like kind of like custodev coach, you know, tell me where I slipped into pitch mode. Tell me where I violated the rules of the Mom Test." The Mom Test has been around for long enough that the agents are kind of aware of the rules in the process. So you can be like, "Hey, like where did I I slip off track from the Mom Test?" I'm also in the process of building a skill set that you can put into your agent, not to do the custodev for you, but to help you kind of analyze what you're doing and ensure that you're following the rules and figuring out where you go off track. Very helpful as a learning tool.
I did a lot of watching my own like recordings or going in with a friend and we would we would do each other's custodev interviews even though we weren't in each other's companies. It was just another founder and like he would come to my some of my interviews and I would go to some of his and then we would take notes on each other and we would compare. It helped me a lot early on. Of course, I didn't have a book like The Mom Test to read, so I had to like stumble my way through it a lot more, but yeah, it's great. Anything to accelerate your learning is a great use of it.
And they're really are like it's stuff that you can pick up from the transcript. You can totally pick it up yourself. Like, if you just read through your own transcript or your own notes, you'll be like, "Whoops, screwed up there. The next 10 minutes of them talking was pure bias, and I set it up when I said this exact thing." You can learn a lot. But yeah, if you find it helpful to to have the agent act as your kind of custodev coach, totally useful.
And then I do think there's some opportunities, not necessarily in the pure customer learning, but certainly like AI facilitated video-based equivalents to surveys and market research for existing customers or for more like demographic research. It's unclear how valuable that is and how much tooling it's worth because surveys are always kind of a niche thing. But there probably are some opportunities there where you could get higher fidelity and better like feature requests and testimonials at scale from existing customers. Practice thing would be cool there.
If if you want to go the practice route for what you're thinking about in terms of your tooling, check out some of the tooling that's around training salespeople. There's a lot of like automated ways to look at the video recordings and the transcripts and basically like get a score even live about how well you're doing as a salesperson. you know, like the percentage of the time you're talking versus listening, like the tone of the conversation. It's like, are you on script or off script? There's like a really rigorous process for basically training up new salespeople within a sales organization. And they've got some pretty good both process and tooling around that. So, if you wanted to borrow some of their playbook and apply it to learning and analyzing your your custodev progress, you totally could. And that would be quite an an interesting thing. I think I think a lot of students and new entrepreneurs would would benefit from from that sort of thing. Not sure how much you could sell it for. because customer learning isn't quite as high value in terms of people's willingness to pay as sales.