What an AI opportunity review should decide

What is an AI opportunity review for?

An opportunity review is not a tool demo and not a strategy deck. It decides whether a specific workflow should be automated, prepared, or left alone. The output is a decision a founder or operations leader can fund, defer, or decline.

Which questions does the review have to answer?

If these are still open, the review is not finished:

1. Which workflow? Name the work, the role that does it, and how often it happens.
2. What does it cost now? Time, delay, rework, or a decision that waits on a person.
3. Where do the inputs live? The systems of record, not a sample file prepared for the meeting.
4. What happens when the output is wrong? Who reviews it, and what the customer or the record looks like if nobody does.
5. What will you measure in 90 days? One baseline, not a list of possible benefits.

A workflow with no baseline is a pilot that cannot succeed or fail. It can only continue.

What are the three legitimate outcomes?

• Build or integrate. The workflow is clear, the data is reachable, and a person is accountable for the result.
• Prepare. The value is real, but the process, the data, or the owner is not ready. Preparation is the project.
• Leave it. The work is rare, the risk of a wrong output is high, or a simpler change in the existing system would do more.

Leaving a workflow alone is a successful review. It saves the budget for the one that should move.

What should you refuse to start with?

Refuse a catalogue of use cases with no rank. Refuse a pilot whose success is "the team liked the demo." Refuse an AI step on top of a spreadsheet that is still the only system of record, unless fixing that system is part of the same plan.

Checklist

• One workflow named
• Current cost or delay written down
• Systems of record identified
• Human review point named where a wrong output matters
• One of three outcomes chosen: build, prepare, or leave it

---