The chat box is not the product

Every product demo this year has the same scene: existing software, new sidebar, blinking cursor, 'ask me anything'. We've decided the interface for intelligence is a text box, mostly because it was the easiest thing to ship.

Chat is a fine interface for conversation and a poor one for work. Work has state. It has steps that depend on other steps, decisions someone is accountable for, and outcomes that are wrong in expensive ways. A transcript is the worst possible record of all of that.

Building Orson, I keep landing on the same questions, and none of them are about conversation. Which decisions will a person actually delegate? How does the interface show what an agent did, what it's about to do, and how to stop it? What does confidence look like on screen when the system is guessing? These are interaction design problems, the same family as basket logic and age verification on a kiosk, just with a probabilistic engine underneath.

An agent acting on your behalf needs terms, and the interface is where they're written. Chat makes those terms vague by default. The AI-native products that last will be the ones that design the workflow first and let conversation be one input among several, the way search became a feature instead of a destination.