Most teams asking whether they're "ready for AI" are asking the wrong question. Readiness isn't about your cloud setup, your data warehouse, or whether you have a Chief AI Officer. It's about whether you can answer five specific questions about the workflow you want to improve.
1. You can name the workflow, specifically.
Not "customer service" — "triaging inbound support emails for our enterprise tier". The more bounded, the better. If you can't draw a box around the workflow on a whiteboard, you can't measure what a model does to it.
2. You know who currently does it, and how long it takes.
Pick three recent examples. Who handled each? How long did they spend? What did "done" look like? If you can't answer in specifics, you don't have a baseline — and without a baseline, there's no way to claim a pilot worked.
3. You have a domain expert who'll label 50 examples.
The single biggest predictor of a successful pilot is whether someone senior will sit down and label 50 real examples with what the right answer should have been. Not "approve a dashboard". Actually label. If nobody will do this, the project will fail — not at the model, at the evaluation.
4. You can describe the failure mode that would matter most.
Every system makes mistakes. A good pilot names the worst-case mistake in advance — the one that would cost a customer, a regulator, or a deal — and designs a human-in-the-loop gate specifically for it. If you can't describe that failure, you're not ready to ship.
5. You know who operates it after we leave.
The handoff is the product. If there's no named owner on your side who'll run the system once it's live — monitor it, update the eval set, review the quarterly report — you're buying a toy. Budget the ownership, not just the build.
Five questions. If you can answer all five clearly, you're ready. If you can't answer three or more, the next engagement isn't a pilot — it's the two-week readiness assessment that turns those blanks into answers.