The verification gap

Every autonomous agent has to answer the same three questions before it acts:

What does this data mean?
Is it current and trustworthy?
What policy governs how it can be used?

This survey helps understand how confident decision-makers actually are in those answers, and the results describe an organization that has built the habit of checking without yet building the certainty that the checks are complete.

87% of decision-makers say they regularly re-verify that the context their autonomous agents rely on is still accurate and current, and 84% say they have clearly defined who is accountable when an agent produces a flawed or harmful output. Those numbers suggest that governance discipline is genuinely being practiced rather than simply claimed.

Source: Collibra/Harris Poll survey, p.5 (Q3)

39% say they have trouble verifying the quality or security of the context their agents draw from. More than a third of decision-makers, in other words, are re-verifying data they were never fully confident in to begin with.

And 72% agree that when their organization’s AI initiatives fall short, the root cause almost always traces back to an unaligned or poor data foundation. Among decision-makers at organizations with $100 million or more in revenue, 96% agree, compared with 81% of decision-makers at organizations with less than $100 million in revenue.

This is, in effect, the ontology problem restated in survey form: A verification process built on a definition, an ownership record or a policy that is itself unreliable will simply produce confident answers that happen to be wrong. Re-verifying more often does not solve that. Governing the context those checks run against does.

That is a catalog and lineage problem before it is an AI problem. An agent can only inherit the trust that the context it retrieves already carries, which means the verification must happen upstream in the definitions and policies themselves, not downstream in a review queue.

72% of decision-makers say AI shortfalls trace back to unaligned or poor data foundation. Among decision-makers at organizations with $100 million or more in revenue, 81% agree.