Only 7% of enterprises say their data is completely ready for AI.

Your Data Is Not Ready for AI explains why, and the explanation surprises most organisations. Asked what obstructs them, enterprises named siloed data and integration difficulty first, at 56%. The absence of a clear data strategy came second, at 44%. Data quality and bias came third, at 41%. Eighty per cent reported limited data access as a constraint. The sharper finding sits elsewhere entirely. Half of organisations have already experienced an AI data exposure incident, where a tool surfaced material the asker should never have seen. Only 34% have formal standards governing how agents access company content.

The paper argues that “is our data good enough” is unanswerable, because readiness is not an attribute of the data itself. It is a relationship between a dataset and a particular purpose. In its place it offers four questions, in priority order: lawfulness, reachability, permissions, and fitness for the specific job. Three of those four have nothing whatever to do with data quality. The first can terminate a programme outright, and organisations characteristically ask it last. Run the assessment against one named use case, and allow the verdict to be not yet.