Your next enterprise evaluation will be run by an agent. Will your signup survive it?

Product-led growth is built on a bet: get the evaluator into the product fast, and the product sells itself. The bet still holds — but the evaluator is changing. Buyers now dispatch agents to build the shortlist: compare pricing, check integrations, attempt the signup, poke the docs. The human shows up later, to a decision that's already half-made.

The thirty-second failure

Run an agent against a typical SaaS signup and watch it die early: an email field with no label or autocomplete token, a "Continue" button that's a styled div, a plan picker whose prices render client-side only, a docs site behind search that agents can't operate. Each one is trivial. Together they mean the agent's report back to its human says "couldn't evaluate" — which reads, in a shortlist, exactly like "no."

What an agent-evaluable SaaS looks like

Pricing that exists in machine-readable form, not just a styled table. A signup flow with labeled fields and honest buttons. Docs whose information architecture works from the initial markup. An llms.txt that routes an agent to pricing, signup, docs, and status. None of this degrades the human experience — it's the same hygiene accessibility has been asking for, now with a commercial forcing function.

In PLG, your signup form was always your real homepage. Now it's your first sales call too — and the caller is a machine.

The competitive angle

Here's the uncomfortable math: when an agent evaluates five tools and completes signup on three, the shortlist is those three. You don't get feedback, a bounce notification, or a second chance. The fix is knowable in an afternoon — test your own funnel the way the agent will run it, fix the failure classes it surfaces, and put the test in CI so next quarter's redesign can't quietly undo it.