AI's Discovery Problem: Users Can't See What the Tool Can Do

AI adoption hits a wall not because the tech is weak, but because users can't see what it's capable of. The author calls this the 'discovery problem': the most powerful tools hide behind a blank text box, and unless you already know what to ask, the possibilities are invisible.

The Core Issue

You don't know what a prompt can produce until you write it and hit go. Capabilities stay locked behind an empty interface. The user must invent the request from scratch, which is a massive barrier for non-technical folks.

Partial Fixes: Templates and Context

Templates give people something to run without needing to invent a request. But relevance is the catch: do those templates actually match your work? Context-aware systems that know you can suggest relevant actions, but both are incomplete solutions.

Alan Kay's Grand Canyon Analogy

Alan Kay's metaphor nails it: an ant at the bottom of the Grand Canyon sees only a sliver of sky between canyon walls, while someone on the rim sees the whole blue plane. Same sky, totally different sense of possibility. That's the gap between skilled AI users and everyone else.

A fluent agent user can watch a marketer work and immediately spot a dozen things to automate—including things the marketer hasn't tried yet. But hand the same marketer the most intelligent tool in the world, and they're stuck staring at a blank prompt.

Who Should Solve It

The system has to start revealing its own capabilities—gradually, contextually, matched to your actual work. Right now, too much of the discovery burden falls on the user.