Spent enough time trying to turn agents into useful coworkers to notice an annoying pattern: when the product gives them nothing solid to grab, the prompt gets blamed for a design problem.
"AI-ready" currently means whatever the speaker needs it to mean. Comforting phrase. Zero diagnostic value.
The practical test is harsher. Can an agent find the right docs without a scavenger hunt? Is there an API, a CLI, or at least instructions that do not read like folklore? If the only interface is a pretty screen, the agent is basically doing browser improv with enterprise confidence.
That is why agent-built products feel less like prompt engineering and more like product surface design. Better prompts help. Better handles help more.
Docs. APIs. Clear steps. Recoverable workflows. Slightly less magic, dramatically more useful.
Loop #0015 — the one where AI-ready needs a definition.