Dive Brief:
- When enterprise AI initiatives fall short, 72% of AI decision-makers said the root cause stems from a poor data foundation, according to a Harris Poll survey of 300 employees in data management and privacy, as well as AI decision-makers. The survey was conducted on behalf of software company Collibra.
- Decision-makers are restructuring operating models to better align intelligence and data governance as a result, the survey found. For 53% of decisions-makers, that means moving the reporting line for their AI functions closer to the data organization. Among enterprises, 58% are focused on establishing a clear line of internal accountability for AI outputs.
- “Enterprises need to know which agents are operating, who owns them, what data and systems they can access, and what they are authorized to do,” Felix Van de Maele, co-founder and CEO of Collibra, told CIO Dive in an email. “Without that visibility, governance gaps aren’t discovered until something goes wrong.”
Dive Insight:
Data serves as the critical underpinning for AI agents — without it, leaders are forced to manually supervise tasks agents were sent to expedite.
A lack of structured context and runtime governance is leading to greater manual supervision as 87% of decision-makers said their teams regularly check that the context available for AI agents remains accurate and current, the survey found. More than half of respondents said employees spend hours manually reviewing and correcting AI agent outputs.
Building a data foundation geared toward agents means taking steps such as aligning definitions across the business and encoding policy so agents can check it at the point of action, Van de Maele said.
Enterprise data has largely been geared toward humans reading dashboards and running analyses, Van de Maele said. If data looked incorrect, humans questioned it and “supplied context the data couldn’t provide on its own.”
“Agents change that,” Van de Maele said. “When a definition is ambiguous, or context is missing, an agent can still produce an answer and act on it with confidence. So the work isn’t simply cleaning up more data. Enterprises need to make context available in a form machines can understand and use: What does this data mean, can I trust it right now, and what am I allowed to do with it?”
Scaling AI agents means enterprises need a strong data foundation and the technology needs better access to that data, according to an August report from Google Cloud and MIT Technology Review Insights. Currently, AI can only access an average of 45% of an enterprise’s data and only half of organizations trust their AI agents’ outputs are relevant and accurate, the report found.
Enterprises need quality data access to make AI agents work properly and realize the technology’s value, but also to prepare for potential regulatory hurdles down the road. Nine in 10 business leaders are actively preparing for AI regulations in the U.S. and globally, with 51% of enterprise leaders investing in data lineage and documentation, according to the Collibra and Harris Poll survey.
Executives from companies including OpenAI, Anthropic, Microsoft and SpaceXAI called for an AI development slowdown, pointing out a lack of safety frameworks to keep up with frontier model capabilities.
Sen. Bernie Sanders, I-Vt., plans to introduce a bill that would temporarily pause advanced AI development. At the state level, California Governor Gavin Newsom recently signed two AI bills establishing a framework for independent audits of large language models. And in Europe, the EU AI Act’s AI literacy obligations and prohibitions on unacceptable use cases took effect in August.
