The AI readiness assessment

Enthusiasm for AI tells you nothing about readiness for it. The point of a readiness assessment is to surface the gaps that will sink a project — while they're still cheap to fix. I score five dimensions, and a weak one anywhere is worth more attention than a strong model everywhere.

1. Data

Is the data that the use case depends on accessible, reasonably clean, and governed? Can you get to it without a three-month access request? Most readiness problems live here. If the data isn't ready, nothing downstream matters yet.

2. Use case clarity

Can you state the problem, the user, and the measure of success in three sentences? “Use AI in customer service” is not a use case. “Cut average ticket-handling time by drafting replies for agents to approve” is. Vague use cases produce vague systems.

3. Ownership and sponsorship

Is there an executive sponsor who wants this to succeed, and a single owner accountable for the outcome? Technology projects without a champion die quietly the first time priorities shift.

4. Skills and operating model

Who runs this once it's live? AI in production needs monitoring, evaluation, and iteration — it isn't a project you finish, it's a system you operate. If there's no plan for who keeps it healthy, you're building a future orphan.

5. Risk and governance posture

How regulated is the domain, and how much explainability will the business and its auditors demand? Knowing this on day one shapes every architectural choice. Discovering it on day ninety forces a rebuild.

Score these honestly and the project plan writes itself — you invest where you're weak before you build where you're strong. It's the cheapest insurance in the whole process, and it's the second stage of my framework for exactly that reason.