AI readiness

AI readiness is usually discussed as if it were about tools, skills or ambition. For a mid-market company, the binding question is narrower and harder: can our data support the AI we are about to pay for?

That breaks into two parts. First, can the data be trusted: are the records correct, filled in, owned, consistent, connected between systems and traceable to a source? Second, can it run the particular use case: does a churn model have one customer ID across CRM and billing, does a cash forecast have payment terms in a system rather than a spreadsheet?

Readiness is always relative to a use case. The same data can be ready for lead scoring and not ready for revenue forecasting. A company is never simply “AI ready”; it is ready for some plays, nearly ready for others, and not ready for the rest.

Readiness also hides between systems. Each system can look clean on its own and still disagree with the others: a customer active in the CRM with no billing account, a closed-won deal with no matching invoice, an account owned by someone who left last year. Those gaps are invisible until records are compared across systems.

8% of businesses say their data is fully ready to support AI at scale (Dun & Bradstreet AI Momentum survey of 10,000 businesses, November 2026). Self-reported readiness is a starting point. Measured readiness, from the company’s own exports, is what a budget decision should rest on.

Two longer reads: what a failed churn model teaches about readiness, and why vendor tools rarely fix the underlying data.