A survey of 300 U.S. data management and privacy employees, and AI decision-makers found that enterprise AI initiatives were falling short because of poor data foundations, according to a CIO Dive report. 

The survey was conducted by Harris Poll on behalf of Collibra, a data intelligence software company. Results were reported on September 17. 

It found that 72% of AI decision-makers at the director level or higher trace their AI initiatives’ shortfalls back to poor data foundations. 

In response, companies are changing how they manage AI. Fifty-eight percent are working to establish clear internal accountability for their AI systems and outputs. And for 53% of decision-makers, that means moving their AI function’s reporting line closer to the data organization.

What the Survey Found

According to the survey report, the findings reflect broader industry trends across enterprise technology. 

A Gartner research report found that at least 50% of generative AI projects were abandoned after proof of concept due to poor data quality, inadequate risk controls, escalating costs or unclear business value.

Collibra co-founder and CEO Felix Van de Maele explained the stakes to CIO Dive. He said enterprises need to know which agents are running, who owns them, and what systems those agents can touch. Without that visibility, he said, gaps in oversight only surface after something has already gone wrong.

The survey found that 87% of decision-makers said their teams regularly check whether AI agents have accurate, current context. More than half said employees spend hours each week correcting AI agent outputs by hand.

This is not the only report to flag the problem. 

An August report from Google Cloud and MIT Technology Review Insights found that AI systems can access only 45% of enterprise data on average. That report also found that only about half of surveyed executives trust that their organization’s AI agents produce accurate, relevant outputs.

Regulatory pressure is adding to the urgency. The same survey found that nine in 10 business leaders are actively preparing for new AI rules in the U.S. and globally, and 51% are investing in data lineage and documentation to get ready. 

Importance of a Solid Data Foundation

Enterprises that rush to deploy AI agents before fixing their data foundations are likely to see the same shortfall repeat. It is important that the data’s meaning, trustworthiness, and permissions are clear enough for a machine to read and act on.

This is where AI governance comes in. AI governance in this context means setting clear rules for who owns each AI system, what data it can reach, and what actions it is allowed to take on its own. 

A few concrete lessons stand out:

• Ownership must be explicit. Among respondents, 58% say their enterprises are still working out who is accountable for what an AI agent does and touches. 
• AI agents need to be provided current or trustworthy data.
• With nine in 10 leaders already preparing for new AI rules, delaying investment in data lineage and documentation adds regulatory risk on top of operational risk.
• The heavy human involvement in correcting AI outputs signals that the problem lies with the underlying data foundation.