Your customers probably don't care about AI

The other day I went to the gym for the first time in a really, really, really long time. (This is the part where I tell myself I'm getting back in shape. We'll see.)

We were talking with the guy who works there about his crappy gym management system and you know what didn't come up? AI! He could not care less about AI. He doesn't think about it. He doesn't follow what shipped this week. He just cared that the system he was using was crappy.

I live in a bubble where everyone is talking about the latest Claude and OpenAI models (okay some of us talk about Gemini models too), looking at how agents are going to change the world. But the truth is most people are like the guy at the gym: completely indifferent to the thing I and many others spend our lives obsessing over (maybe it's job security).

I love what is happening with AI. I use it in everything I work on. But when I work on products, or at least most products, the customer is usually like the guy at the gym, and they do not care whether AI is part of the solution or not. The only thing they care about is whether it solves a problem they are having and how well it solves it.

Somewhere in the last two years, "AI-powered", "agentic", etc. became the feature. They're on the landing page, in the sales deck, in the first sentence of every pitch. And most of the time it means a team fell in love with a technology instead of the customer.

Think about what your customer actually asked for. Has anyone actually ever opened your product and thought, "I hope there's an agent behind this button"? They thought: I want this report done before my 4pm. I want to not get yelled at by my boss. I want ... The AI is your business, not theirs.

I have spent most of my working life in the fraud, risk and security space, and let me tell you something I know. On the products I dealt with, the end user cares that they don't lose money to fraud, that their good customers are not blocked and that their information is safe. If I solve that with a model, great, if I use a fancy agentic solution great and if I solve it with three boring rules and a velocity check, also great. The customer can't tell the difference and shouldn't have to. The moment I start selling the model instead of the outcome, I've confused my problem with theirs.

We should use AI, we should use it everywhere it makes someone's life better. Use it to automate tasks, to catch the thing a human would've missed, to answer customers faster, to delete a step the customer always hated. Just use it the way you use a database or a load balancer: in service of the experience, not as the experience. In most cases, the best AI in your product is the AI your customer never notices, because they're too busy getting what they came for.

When a technology is genuinely good, it becomes like plumbing. You don't see it even if know it is there, and it makes your life better.

There is one real exception. If you are an AI company, if the model is the product, then yes, your customer cares deeply about the AI and you should talk about it endlessly. Anthropic should talk about AI. OpenAI should. Their customer is literally buying AI.

But you are probably not that company. Most of us aren't.

Now excuse me while I go back to building Boring Ops with Claude and Codex.