user experience > AI model
for once, builders are more flaky than users
AI injects itself into every product conversation. It has to. We’re so early in the Age of AI, but we’re already Post-AI. It’s here, so we might as well build and play with it.
As you build with it, you may notice something:
The back end doesn’t matter as long as it delivers a good experience to your users.
Users don’t care which LLM delivers insights. If the insights are helpful, users are happy.
Greg’s tweet made me reflect on how we’re building our customer health scoring insights in Accoil. We’re using AI to act as your team’s product data analyst. It’s always on, hunting for signs that users may cancel their subscription or for accounts that may be ready to upgrade.
We decided to give users the choice of model that drives their own insights engine. OpenAI, Claude, Perplexity, they’re all in there.
Not many people care
Turns out that those of us building the products are a lot more obsessed with the underlying tools. Of all the conversations I’ve had about an AI-driven insights engine, only a small minority of them ever touch on “which model are you using?”
Greg is right:
Customers and users aren’t that interested in the AI models. They want (and they use) tools that give them a predictable user experience.
That is up to you and I to build. We can swap out the AI engine any time. In fact, I think it’s up to us to do the work to know which models are best for our customers.
It’s also up to us to deliver the best product experience possible. Think of it like buying a car.
Some people care a lot about the size of the engine, the fuel economy, predictable service schedules. But most people want to push a pedal and go vroom in a car that makes them feel good about themselves.
As Greg says, “technology advantage is temporary. Interface lock-in is forever.”
Build with your user experience in mind and remember that all the AI models are here to help you serve.
Peter
(368 / 500)