A short talk by Mike Lyons arguing that AI should be used as a practical tool to help teams understand customers faster and at scale. He outlines common failures (misunderstood customers, invisible motivations, and scale limits), gives a playbook for building customer personas from messy data using language models, and warns about persona drift, over-segmentation, and AI bias.
A few months back, I was sitting in a conference room watching a team demo. A feature that they'd spent 12 weeks building for this major retirement system. The team was excited. I mean, you could see it on their faces. They demoed and they thought they just nailed it. But at the end of the demo, there was just silence in the room. Well, that's embarrassing. They'd spent three months solving the wrong problem because they never really understood their customers. Look, this happens more than you'd think. Teams build features based on what they think customers need, not what customers actually need. Now, the problem is not that these teams don't care about their customers. Of course they care about their customers. The problem is that understanding customers well, it's harder than it looks. Well, this is where AI comes in. Not as some silver bullet that's going to solve all your problems, but as a tool to help you understand customers better. AI can read through thousands of customer comments and support tickets and reviews in just minutes, as opposed to like weeks. And it can spot patterns across different customer segments that be really difficult to catch manually. It can help you ask better questions about your data that you might not even think to ask. And it can then turn that data into insights that actually matter and help us be more successful. Well, let's think about some of the things that most companies face on this topic, and maybe this is true in your organization. So see if this resonates with you. First, lots of companies, well, they don't know who their customers really are. Sure, they might have basic demographics like age and location, but they don't understand the deeper stuff, like what motivates their customers, what frustrates them. Well, AI can analyze unstructured feedback to reveal motivations and frustrations that might otherwise be hidden in mounds of data. Now, just considering our case study at Technova, for example, they know their task flow users while they work at creative agencies. But do they understand the difference between a project manager at maybe a 10 person design studio versus someone at a 50 person marketing agency? Well, see, those are completely different people and they have different needs. The second challenge is that customer behavior is, generally speaking, invisible. Sure, you can see what your customers do, but you can't see why they do it. You might notice customers stop using a feature after a couple weeks, but you don't know if it's because the feature is confusing or unnecessary or just kind of buried deep in the interface. Well, AI can connect the dots between what customers do and why they do it by analyzing patterns across multiple data sources. And the third challenge is scale. Even if you did want to understand your customers better, how do you collect data and analyze feedback from hundreds or thousands of people? Heck, by the time you've gathered enough feedback and manually gone through it to spot patterns, the whole market might have shifted. But AI can process feedback from thousands of customers in real time, turning overwhelming data into clear insights. Now, a few years back, I was working with a marketing team, and they were convinced that they had one type of customer. They called them busy professionals who need to stay organized. Well, their entire product strategy was built around this single customer Persona. Every feature, every marketing message, every pricing decision was designed for these busy professionals. But when they fed their customer data into AI, support tickets, usage patterns, survey responses, it immediately identified four distinct groups with completely different needs and behaviors. Yeah, they're all busy professionals. But seeing these distinct Personas let them build features that actually matched how people use their product. All right, you might be wondering, so how do you actually build customer Personas using AI? Here's the playbook I use with teams. We're going to start by gathering everything you've got. Pull together support, tickets and survey responses, product usage data, online reviews, basically any place your customers are telling you what they think. Now, my favorite part about all this is that you don't really have to worry about the data being messy. AI loves messy data, but let it do the heavy lifting. Feed all that data into your language model and ask it to group customers by the themes that keep coming up. You'll start seeing patterns like overwhelmed by too many tools or struggling with remote team coordination. But give each group a name that makes sense. Don't just call them segment A, B, C. Give them a unique name. Call them something memorable like multi tool users or remote team managers. Now, you want to capture that. Keep it simple. One page per Persona. Their goals and frustrations, how they behave. Keep it readable. You want your whole team to actually use these? Then, then, and this is a crucial piece. Check your work with real people. Show those Personas to sales and support and most importantly, to customers themselves. Do they see themselves in this description? If not, go back and fix them. Because here's the thing. AI speeds up the pattern finding and data crunching, but humans, well, we still need the sanity check. We still need to do the storytelling and the checking in. Do these seem right now, just a few things to watch out for as we kind of close out here. Persona drift. It happens. These aren't static documents. Your market changes, your customers evolve. So run this analysis every once in a while, maybe every quarter. And update your Personas. Also, don't slice these too thin. I see teams create 10 different Personas when four solid ones would be way more useful. More isn't always better. And remember, AI has blind spots. It learns from historical data, which means it can learn historical biases, too. So always look at what AI gives you with a critical eye. The companies that are winning today, they're not the ones with the best guesses about their customers. They're they're the ones with the best insights. And customer Personas. Well, they're the foundation of every product decision that you make. AI helps product managers process massive amounts of customer data quickly and turns those insights into a competitive advantage. So the question isn't whether you should use AI for customer insights, it's how quickly can you start.
