The AI Readiness Checklist for SMBs.

What the four sections are actually testing

The questions look like a maturity assessment. They're really a filter for the four failure modes that account for most stalled AI projects — and a low score in one area matters far more than an average score across all of them.

Data — is the answer reachable?

Not whether you have a warehouse, but whether the specific facts a system would need to answer correctly exist somewhere a machine can read. Policies in a PDF nobody has opened since 2021 score badly here, and this is where most promising projects quietly die.

Use cases — is it repetitive and checkable?

High volume gives you signal to tune against; a clear notion of correct gives you something to tune toward. A task that happens twice a month, or where nobody can say whether an output is right, isn't a first project regardless of how appealing it sounds.

Governance — who says no?

Someone has to own which actions require human approval, what happens to customer data, and who is accountable when the system is wrong. Organisations that can't name that person don't fail at the build — they fail at the launch review, months later.

Enablement — who owns it in six months?

An AI system needs someone to review escalations, extend the eval suite, and adjust prompts as the business changes. If no named person has time for that, the system degrades quietly until people stop trusting it — which looks like a technology failure and isn't one.

Scoring low on data or governance usually means the first engagement should be strategy rather than a build. That's a cheaper conversation to have now than after a failed pilot.