A major new research report from Veeam has put a number on something many business leaders have been sensing for months: most companies are running AI agents they do not understand, cannot govern, and are not equipped to manage.

The report, titled “AI’s Promise is Colliding with a Data and AI Trust Gap,” surveyed 600 senior executives across financial services, healthcare, manufacturing, retail, and technology from North America, Europe, and Asia-Pacific. The findings are striking.

Eighty-eight percent of organizations are already using or piloting AI agents. But only 7 percent qualify as “truly AI-ready.” That gap is not a small misstep. It is a companywide vulnerability.

The Numbers Behind the Problem

Three statistics from the Veeam report deserve to sit in your head for a while.

The first: 95 percent of organizations said data challenges have already slowed their AI progress. Not might slow it. Already have. In almost every organization running AI, something about how their data is structured, stored, or governed is creating friction.

The second: only 28 percent of organizations are confident they can detect an AI system operating outside approved parameters. That means nearly three-quarters of companies deploying AI agents have no reliable way of knowing when those agents are doing something they should not be doing.

The third: only 29 percent of organizations can identify within minutes which systems an AI agent has accessed. For 22 percent, tracing which data was used by an AI is anything close to fast.

These are not theoretical risks. These are operational blind spots inside companies that have already deployed agents.

Why This Is Happening

The pattern Veeam describes is one most business leaders will recognise once they see it named clearly. AI adoption moved faster than the infrastructure designed to support it.

When ChatGPT arrived in late 2022, companies started experimenting. When large language models became capable enough to automate real tasks in 2024, companies started deploying. When agentic AI tools became available through platforms like Microsoft Copilot, Salesforce Einstein, and ServiceNow, companies started scaling.

At each stage, the urgency to deploy outran the discipline to govern.

Data quality is often the first casualty. AI agents are only as reliable as the data they operate on. If customer records are incomplete, product information is inconsistent, or financial data exists in disconnected silos, the agents will either produce wrong outputs or fail silently. Most companies discover this after deployment, not before.

Access controls and audit trails are usually the second. Knowing what an agent did, what it touched, and whether it behaved as intended requires logging infrastructure that many organizations built reactively, if at all.

What 7% AI-Ready Looks Like

The 7 percent of organizations Veeam identifies as truly AI-ready have something in common. They treated data infrastructure as a prerequisite to AI deployment, not a parallel track or an afterthought.

They also invested in governance structures before rolling out agents at scale. That means knowing who owns each AI system, what data it is permitted to access, what it is allowed to do, and how anomalies get flagged.

This is not glamorous work. It is the kind of foundational investment that does not make a press release. But it is what separates organizations that benefit from AI from organizations that are exposed by it.

What This Means for Business

If you are a business leader, this report is telling you something important: the risk of AI is not that it will not work. The risk is that it will work, and you will not know what it did.

Three things matter right now.

Know what you have deployed. Many organizations have more AI running than their executive teams realise. Copilot integrations, automated support tools, process automation, and data pipelines with AI components can accumulate quickly. A full inventory of what is deployed, what data it accesses, and what decisions it influences is the starting point.

Fix your data before you scale your agents. Garbage in, garbage out still applies. The difference is that when a human analyst produces a wrong output, it usually affects one decision. When an AI agent produces a wrong output and it is embedded in a workflow, it can affect thousands. Data quality investment is agent-readiness investment.

Build for observability. You need to know, in close to real time, what your AI systems are doing. This means logging, audit trails, anomaly detection, and clear escalation paths when something looks wrong. If you cannot currently answer “what did my AI agent do in the last hour,” that is a gap worth addressing before you expand deployment.

The Veeam research captures a moment in enterprise AI that will define the next two years. Companies that close the gap between adoption and readiness will build competitive advantages that are genuinely durable. Companies that do not will accumulate operational risk until it surfaces in a way they cannot ignore.

The gap is real. The question is which side of it you intend to be on.

---

Enterprise DNA helps organisations build the data foundations, upskill the teams, and implement the AI strategies required to move from AI experimentation to AI readiness. Learn more about Omni Advisory or explore EDNA Learn for team upskilling.