Enthusiasm for agentic AI is near-universal in enterprise boardrooms. Actual deployment at scale? That’s a much smaller club.

A new study commissioned by Teradata and conducted by Wakefield Research surveyed 1,000 senior technology and data leaders across six global markets between March and April 2026. The headline finding: only 7% of enterprises have reached what the researchers call the operational stage, where agentic AI delivers tangible business outcomes.

The report is titled “Arrested Automation: Why Agentic AI Stalls at the Enterprise Level” and it lands at a revealing moment for an industry that has spent two years talking about AI agents as the next big thing.

The Gap Between Ambition and Reality

The majority of enterprises (68%) remain stuck in the Experimenting or Developing stages of agentic AI maturity. That means they’re running pilots, maybe deploying a few internal tools, but not yet seeing the kind of workflow transformation that justifies the investment.

The report identifies a root cause that will frustrate anyone who expected better models to solve the problem: the data foundation most companies have built over the past decade was not designed for agents.

Specifically:

• 77% of executives report that 20% or less of their enterprise data is sufficiently described and contextualized for AI agents to actually use
• 78% of leaders find it challenging to unify data and knowledge across business functions so agents can reason across the whole enterprise
• Context fragmentation, where data exists but carries no usable meaning for agents, is the primary limiting factor holding back the majority of deployments

In other words: you can have access to the best AI models on the planet, but if your data is locked in siloed systems with no consistent context, the agents cannot do their job.

The Leadership Perception Gap

There is also a meaningful gap in how different levels of an organization perceive where they stand.

69% of C-suite executives say their organization is already operating with agentic AI. Only 57% of VPs say the same thing. The further down you go from the boardroom, the more clearly people see the gap between what was announced and what is actually working.

This kind of perception gap is common in enterprise technology transformations. The executive hears the roadmap pitch. The VP sees the rollout reality. When only 7% have actually crossed the finish line, a lot of confident announcements are still describing future states.

What This Means for Business

For business leaders thinking about their own AI agent strategy, the Teradata findings point to a concrete to-do list before deploying agents at scale:

Get your data house in order first. The report is clear that the constraint is not model capability, it is data readiness. If your enterprise data lacks consistent structure, metadata, and context, agents will produce unreliable outputs. The same discipline that made data warehouses useful in the 2000s applies here: clean, described, governed data is the foundation everything else depends on.

Close the perception gap deliberately. If leadership believes agentic AI is live and working while the people running it know it’s still in pilot mode, that misalignment will produce bad decisions. Regular, honest reporting on what the agents are actually doing (and not yet doing) keeps strategy grounded in reality.

Start with bounded workflows. The 7% who have scaled tend to have something in common: they picked narrow, well-defined processes first, proved measurable results, and expanded from there. The impulse to deploy agents everywhere at once is what typically leads to context fragmentation and stalled outcomes.

Invest in data literacy alongside AI tools. Agents are only as useful as the data and instructions they receive. Teams that understand how to structure data, write good prompts, and interpret agent outputs will get dramatically more value than teams handed a tool with no upskilling. This is not a soft skill; it is increasingly the primary factor separating companies that see ROI from those that don’t.

The 93% of enterprises still working toward full agentic AI scale are not behind because they lack ambition or resources. They’re behind because the underlying data infrastructure needs to catch up with the strategy. The companies that close that gap fastest will have a structural advantage in operational efficiency that compounds over time.

---

Research methodology: Wakefield Research commissioned by Teradata, 1,000 senior technology and data leaders, six global markets, fieldwork conducted March 23 to April 5, 2026.

Enterprise DNA helps organisations build the data foundations and AI capabilities needed to move from pilot to production. Talk to us about your AI readiness strategy.