AI agents outpace trust in enterprise data, survey finds

The Modern Data Company has published interim findings from its annual survey of enterprise data leaders, suggesting AI agents are being adopted faster than companies trust the data behind them.

The survey gathered more than 540 responses from enterprise data leaders and practitioners across 66 countries. It found that 57.3% of respondents are piloting or running AI agents in data and analytics workflows, with 23.5% already using them in production and 33.8% in pilot programmes.

Another 15.6% expect to begin using AI agents within six months, while only 4% have no plans to do so. The figures suggest AI agents have moved quickly from experimentation into operational use across many organisations.

The data foundation behind that shift appears much weaker. Just 8.4% of respondents said the data feeding their AI systems was trustworthy enough for production use.

Even among companies already running AI agents in production, confidence remains limited. Only 21.7% of that group said they were very confident their data was trustworthy enough for production.

Trust gap

When asked what was preventing agents from reaching production, 75.9% of respondents cited data quality and trust as one of the top three barriers. That ranked ahead of missing context and lineage at 63.5% and security concerns at 61.7%.

More commonly cited explanations for slow AI adoption ranked lower. A skills gap was named by 25.9% of respondents, while 19.5% pointed to immature tooling.

The findings suggest companies are constrained less by access to AI systems than by their ability to govern and verify the information those systems use. As AI agents move from assisting analysis to taking actions across systems and workflows, the reliability of underlying data becomes a more immediate operational issue.

Business context emerged as another weak point. The survey defined this as the definitions, relationships, lineage and policy that explain data and how it should be used.

A clear majority of respondents, 60.9%, said a reliable context layer is necessary for AI agents. Yet only 16% said their organisation deliberately designs and engineers that layer as a product, while one in four said they have no formal context layer at all.

Asked where a single investment would make their data and AI work more effective, respondents chose a better context layer over better tools by roughly six to one. Even among organisations that describe themselves as AI-first, only 38.5% said they had engineered their context layer.

Governance shortfall

The survey also pointed to a gap between expectations for explainability and current governance practice. While 65.1% of respondents said AI-enabled decisions must be explainable, traceable and defensible under scrutiny, far fewer said their organisations have the systems to support that standard.

Only 39% said they maintain either an audit trail for AI inputs and outputs or a link from decisions back to the original data sources. Just 10% said they maintain both.

Accountability frameworks also appeared underdeveloped. Only 17.7% of respondents said their organisation has a clear, documented AI accountability framework, while 25.4% described accountability as shared but unclear. Another 19.1% said it was poorly defined, and 11% said no one was formally accountable.

Platform strategy formed another part of the picture. Nearly half of respondents, 47%, said they were actively consolidating towards fewer platforms, while 17.2% said they were evaluating that move.

At the same time, most companies have not fully abandoned specialist tools. Nearly 90% of organisations that are consolidating still rely on a best-of-breed point solution somewhere in their technology stack.

The survey also found differences between companies that have already moved AI agents into production and those still earlier in the process. Organisations already running agents in production were nearly four times as likely to have intentionally developed a context layer and almost three times as likely to trust the data underpinning their AI.

It also found that organisations with an engineered context layer were roughly five times as likely to report they could draw validated causal links between their data work and business outcomes. The survey noted, however, that these findings show correlation rather than causation.

Saurabh Gupta, President and Chief Executive Officer of The Modern Data Company, said the results point to a widening gap between deployment and preparedness. "Enterprises have proven they can put AI agents to work. The harder question is whether those agents have the trusted data and business context they need to operate reliably," Gupta said.

He added that more advanced adopters were also further along in their data work. "The organizations further along with agents are also further along in building that data foundation. That is an important signal for every company trying to move AI into production," Gupta said.

Julia Bardmesser, Chief Executive Officer of Data4Real and a member of the Modern Data 101 community, said the pattern matched what she had seen in the market. "This research matches what I see in practice: AI is moving into production faster than the data supporting it is becoming trustworthy," Bardmesser said.

She said the consequences grow as agents take direct action. "The gap is familiar, but what has changed is its impact. In traditional analytics, questions about the reliability of the data could be addressed before someone acted on the result. With agents, that same data can lead directly to action. Data quality does not have to get worse for the consequences to grow considerably," Bardmesser said.