By 2028, 40% of IT infrastructure and operations teams that expand their use of AI without strong controls will face a disruption to a service their business depends on, according to Gartner Inc., a business and technology insights company.

Gartner surveyed 298 infrastructure and operations (I&O) leaders in the United States and the United Kingdom in June and July 2026. It found that 95% of their organizations had experienced at least one incident involving an AI agent in the previous 12 months. AI agents are software tools that can take actions to complete a task.

“Establishing AI governance that includes policy controls, permission models, observability instrumentation, and audit capabilities is critical,” said Nathan Hill, analyst and chief of Research at Gartner. “However, there is a massive operational readiness gap as 70% of I&O leaders believe their AI agents currently operate within safe boundaries.”

Only 15% of respondents had an automatic way to stop an agent when its errors got out of control or it got stuck repeating the same steps. In the same survey, 57% described their rules and controls for AI as weak, and 55% said the tools they use to manage IT risks were completely outdated.

“I&O leaders must immediately inventory active AI agents and classify them all as nonhuman identities in their identity platforms. AI agents aren’t just another automation script; they require a fundamentally new, proactive infrastructure governance approach to balance productivity and innovation with enterprise safety,” said Hill.

Gartner recommends keeping a list of AI agents, tracking the steps each one takes, and requiring a person to approve actions that carry serious risks. IT teams should be able to slow down an agent or isolate it before an error affects other systems.

The company also recommends setting limits on what an agent can spend during a task, choosing when it should use a private or public AI model, and controlling what data it can access. Gartner analysts are discussing the findings at the Gartner IT Infrastructure, Operations & Cloud Strategies Conference in London.

“As dependence on agents grows, I&O leaders have less flexibility to switch them off. A governance framework must identify which agents are critical to keep running, even if they are expensive or challenging to maintain,” said Hill.