Free 5 criteria about 4 minutes

Score it before you automate it.

An AI agent doesn't fail like software. It fails like a confident intern who says the job is done when it isn't.

These are the five criteria I score every workflow against in the Agent Readiness Audit. Run your own workflows through them below and you'll get a verdict, a score, and a ranked fix-list for each one.

Everything is scored in your browser. Your workflow names and answers are never sent anywhere — there is no request to score, and nothing is stored except in this tab. If you ask me to keep your map, the only thing that leaves is your email address. The map itself is built right here.

What the five criteria mean

01

Documented

Could someone new run this from a written description, without asking anyone?

Agents automate descriptions of work. No description, no agent.

02

Deterministic

Are the judgment calls written as rules, or is it “it depends”?

Every “it depends” is a place the agent guesses — and guesses confidently.

03

Verifiable

Is there an automatic check that the work was done right?

If the only check is “a customer eventually complains,” the agent fails silently for weeks.

04

Recoverable

If the agent gets it wrong, what does undo cost?

Cheap undo, automate freely. Expensive undo, a human stays in the loop. No exceptions.

05

Delegatable

Does a named human own the output, and can the agent earn trust gradually?

You onboard a new hire in stages. Same rule here — nobody gets the keys on day one.

Two of these are hard stops rather than scores. If there is no way to check the work, the agent will fail silently for weeks. If undo is expensive, a human stays in the loop no matter how good the rest looks.