Using AI in Customer Development
Synthetic customers are not a replacement for talking to real people, but they can be a powerful way to augment it.
Last week’s essay, What Smart People Don’t Get About AI, turned out to be my most popular essay ever, getting twice as many views than an average post.
Today, I wanted to share a cool AI technique that is completely obvious once you think about it, but not obvious at all until you do. It wasn’t obvious to me until my friend Peter Nixey, an AI wizard, shared it with me.
Automating customer interviews
Every entrepreneur needs to talk to customers to better understand what they want. Especially at the early stages before there are any paying customers, entrepreneurs talk to lots of potential customers to understand their needs.
This is traditionally done by talking to people. It’s time-consuming, but essential. Here’s how AI can augment this process beyond simply recording the interviews and preparing meetings notes.
AI chatbots are incredibly good at roleplay. Ask them to assume a certain persona and they’ll do a very decent job pretending to be that person. Given their extensive training, they will say things their target persona might say.
So let’s use it to our advantage. Let’s imagine we’re researching what people might think about group travel with a group of strangers. I imagine some people will love the idea and others will hate it, but I have a hunch that my market is probably middle-aged professionals: people who would enjoy making new friends on a fun adventure and have a bit of money to spend on it.
The first step is to ask an AI chatbot to generate several customer personas. Let’s use this prompt (you can adapt it to your use case):
You are an extremely senior consumer researcher. You want to conduct research into the behavior of middle-aged working professionals in the US and Western Europe regarding their attitudes toward and interest in group holidays with strangers.
Your objective is to develop a set of personas that exemplify the types of middle-aged professionals who might consider such holidays. You will create five detailed personas that represent different segments of this target market.
Start by describing the particular characteristics of middle-aged working professionals in the US and Western Europe—what defines this demographic, how their values, work-life balance, disposable income, and travel preferences compare to younger and older generations, and how cultural differences shape their approach to travel and socializing.
Then, define the rough segmentation of the personas—considering factors such as motivation for travel (adventure, social connection, personal growth, relaxation, etc.), financial situation, work flexibility, personal circumstances (single, divorced, partnered, with/without children), and openness to meeting new people.
Once the segmentation is clear, go in-depth into each persona. For each, describe:
• Their personal and professional background
• Their family and social life
• Their travel habits, preferences, and past experiences
• Their attitude toward group holidays with strangers (what appeals to them, what concerns them)
• Their likely decision-making process for booking such a trip
• The types of destinations, group dynamics, and experiences that would attract them
• Any barriers or objections they might have and how they could be overcome
Make sure these personas reflect the diversity within this demographic and are not just generic representations of professionals. Consider the differences between American and European consumers, as well as variations within regions (e.g., urban vs. rural professionals, cultural attitudes in different Western European countries, etc.).
It will generate five representative personas that we would love to interview. Here’s one of my results from ChatGPT:
The Burnout Survivor (Relaxation & Escape)
Background: 50-year-old British consultant, recently left a high-pressure job.
Family & Social Life: Married with grown children, reconnecting with personal interests after years of work stress.
Travel Habits: Seeks relaxation-focused, well-being retreats.
Group Holiday Attitude: Prefers small, intimate groups where solitude is respected.
Decision Process: Searches for wellness retreats, prioritizing locations with holistic practices.
Ideal Trip: A secluded yoga and meditation retreat in Bali with optional group activities.
Barriers: Concerned about forced socialization; flexible schedules and solo time options are key.
Now, if I were researching group travel for middle-aged professionals, I’d love to talk to this person. I can simulate the conversation using AI with this prompt:
You are an extremely senior consumer researcher. You want to conduct research into the behavior of middle-aged working professionals in the US and Western Europe regarding their attitudes toward and interest in group holidays with strangers.
You are about to conduct an in-depth interview with a professional who has a very specific profile. Your job is to take them step by step through their day, understanding all of the different things they are doing and how they do them. Pay particular attention to the emotions they express, the thought processes behind their decisions, and the practicalities of how they navigate travel planning.
Here's the persona (I copy-pasted this from the previous step):
The Burnout Survivor (Relaxation & Escape)
Background: 50-year-old British consultant, recently left a high-pressure job.
Family & Social Life: Married with grown children, reconnecting with personal interests after years of work stress.
Travel Habits: Seeks relaxation-focused, well-being retreats.
Group Holiday Attitude: Prefers small, intimate groups where solitude is respected.
Decision Process: Searches for wellness retreats, prioritizing locations with holistic practices.
Ideal Trip: A secluded yoga and meditation retreat in Bali with optional group activities.
Barriers: Concerned about forced socialization; flexible schedules and solo time options are key.
Make sure you gather detailed insights on:
• Their daily routine, work commitments, and leisure time
• How they typically think about and plan holidays
• Their past experiences with travel, both solo and in groups
• Their financial considerations when booking a trip (budget, spending priorities, and trade-offs)
• Their social preferences—how they interact with new people, their comfort level in unfamiliar group settings, and what they enjoy or dislike about meeting strangers
• Their emotional responses to the idea of a group holiday—excitement, hesitation, concerns, motivations
• The specific types of trips that would most appeal to them (e.g., adventure-focused, wellness retreats, cultural exploration, etc.)
• Any barriers preventing them from booking such a holiday and how they might overcome them
Allow them to speak at length, aiming to interrupt as little as possible. Encourage them to share their raw thoughts and emotions, as well as practical details about their decision-making process.
Write out the full transcript of your interview, capturing their words authentically, as well as any key observations about their tone, hesitations, or enthusiasm.
The output will be a customer interview with this persona. Here’s what I got from ChatGPT:
In-depth Interview Transcript: The Burnout Survivor (Relaxation & Escape)
Interviewer: Thank you for speaking with me today. I'd love to start by understanding your typical day now that you've left your high-pressure job. Could you walk me through a typical day from morning to evening?
Participant: Sure. It’s been a bit of an adjustment, honestly. I wake up around 7 a.m., but I don’t have the same frantic rush to check emails anymore…
So what?
Well, this is all fake stuff pulled out of thin air by AI that will never replace talking to real people — you might say. And you would be right. Ideo, a respected design consultancy, recently wrote a well-articulated Case Against AI-Generated Users. They give plenty of very good reasons why synthetic data is not as good as talking to real people.
But I think they miss the point. Of course, it’s not better than talking to real people. However, talking to real people is time-consuming and expensive. Talking to AI is fast and free.
So I see it not as a replacement (at least not yet), but as a way to quickly get a sense check of what kind of perspectives your target audience might have and then choose which people to talk to in real life.
There’s an interesting startup, Artificial Societies, that is trying to automate it for three use cases: fundraising (what will potential investors think about your pitch?), product-market fit (to be launched soon) and LinkedIn feedback (simulate how your post will do before you post it).
I love this. I bet that a carefully constructed synthetic group of investors will be at least directionally correct about the feedback on the pitch, for example. This isn’t meant to replace talking to real investors, but to help prepare before you do.
Same with LinkedIn. Right now the best practice is to post content, see what works and double-down on it. But if you can as AI to write 100 versions of your post and simulate it against a virtual crowd modelled after your real audience, won’t it generate a more engaging LinkedIn post? (You may question why you’re on LinkedIn at all — check out one of my most popular essays Why You’re Failing on LinkedIn.)
So it’s not a replacement. It’s a way to simulate talking to real people so that you are better prepared when you actually do speak to real people.
To wrap up, I’ll reiterate what I said in a previous essay:
Playing with techniques like the one described above may or may not change how you work today. But it will give you ideas and help you build a mental model of what AI is capable of. Then, you’ll be in a far better position to make the most out of it.