AI Failures Often Trace Back to Poor Data Foundation: Survey
CIO Dive, Thursday, September 17th, 2026
87% of decision-makers say teams regularly check agent context is still accurate, and half spend hours correcting outputs.
A Collibra study conducted by Harris Poll found more than half of enterprises working to establish clear internal accountability for AI outputs, while a lack of structured context and runtime governance drives heavy manual supervision.
Some 87% of decision-makers said their teams regularly check that the context available to AI agents remains accurate and current, and more than half said employees spend hours manually reviewing and correcting agent outputs.
Collibra's Felix Van de Maele argues enterprise data was built for humans reading dashboards, who questioned figures that looked wrong and supplied context the data could not provide; an agent facing an ambiguous definition still produces an answer and acts on it confidently.