Essay · May 10, 2026
A counterintuitive case for incumbents ceding discovery to AI agents — and owning what's left.
In 2007, every record label was fighting to own the store. Five years later the store didn’t matter, and the companies that won were the ones that owned the listening.
We’re about to run that play again, and most incumbents are lined up on the wrong side of it. The fight right now is over discovery — who the customer asks first, whose box they type into, which agent answers “what should I buy.” Everyone’s pouring money into being the front door. It feels obvious. It’s the same instinct the labels had about the store.
It’s the wrong fight.
When an AI agent sits between your customer and your product, the discovery layer commoditizes. The agent asks the question; a dozen providers answer; the customer never sees most of them. Spending to be the one they ask first is spending to win a race that’s getting shorter every quarter. Meanwhile the thing that doesn’t commoditize — the usage, the data your product throws off while someone actually uses it, the relationship after the first click — is sitting there unguarded while everyone fights at the door.
So cede the door. Let the agent have discovery. Make your product the best possible thing for an agent to recommend, then own everything that happens after the recommendation. The moat was never the front door. It was the house.
If you’re an incumbent, here’s the one question worth arguing about in your next strategy offsite: what are we still spending to win at the door — and what would we do with that money if we admitted we’d already lost it?