Full Circle

We spent the last decades trying to replace humans with machines. The process was straightforward: identify the actions humans make and replace them with APIs or deterministic functions that achieve the same outcome.

This required limiting options and having very explicit user interfaces — which worked well for most people, but not all. No matter how good the UX was, there was always something you wanted to do that you couldn’t.

With AI-native apps, we have to think about it the other way around. Now we’re replacing machines with humans — or rather, with AI agents that take the role humans used to play. We’re no longer bounded by specific actions and APIs. We can rely on direct communication and understanding user intent.

Take cashiers at restaurants. We first had humans taking your order, understanding all the details — special requests, substitutions, everything.

Then came automated kiosks. They replaced the cashier with a digital interface, but this required an additional layer of translation. Users had to transform their intent — something as simple as “I want a burger, large fries, no mayo” — into a sequence of taps: select quantities, pick ingredients, type special requests. That was significant effort transferred from the human cashier to the user.

Now with AI agents, we can give that complexity back to the machine. Humans go back to saying “two burgers, two fries, no mayo, extra ketchup” — and the machine translates it into actions.

The interface comes full circle. The complexity shifts back to where it belongs.