Enterprises are racing to deploy agentic AI, but most are not yet ready to manage it. Gartner's latest survey of technology leaders finds that 71% of organizations expect to deploy agentic AI within the next 12 months, while only 21% currently have agentic AI governance capabilities in place. The gap is wide enough that Gartner now warns of service disruptions for companies that scale agents without strong runtime controls.
The Survey Numbers
The headline figures come from Gartner's Q2 2026 survey of more than 2,000 CIOs and technology executives, which the firm published in early November. Coverage of the findings reports the 71% planned-deployment figure and the 21% governance figure, a 50-point gap between ambition and readiness. The survey points to a familiar pattern in enterprise technology: adoption moves quickly, while the controls, ownership and audit practices that keep it safe lag behind.
The gap matters more than either figure alone. A majority of enterprises are planning to move agents into production within a year, but only about one in five has the governance structures to manage them. Reporting on the survey also notes that the figures describe organizations' plans and capabilities rather than the outcomes of those deployments, so the real test will come as agents take on more consequential work.
Runtime Governance Is the Missing Layer
A second Gartner study, presented at its London infrastructure and operations conference on November 16, sharpens the point. It surveyed 298 I&O leaders in the United States and United Kingdom between June and July 2026. Its findings describe what governance looks like in practice when agents are already running:
- 95% of organizations had at least one agent-related incident in the past 12 months.
- 57% rate their AI governance frameworks as weak.
- 55% say their IT risk tools are completely outdated for autonomous machine behavior.
- Only 15% have automated circuit breakers that can halt a runaway agent loop or stop an invalid system change in real time.
- Only 22% have a designated, agent-specific control plane.
Those numbers describe a mismatch. Many leaders believe their agents behave safely, with 70% saying so in the study, yet few have the technical means to stop an agent that goes wrong. Gartner's view is that agents are not simply another automation script. They act continuously and can change systems, so they need controls that operate at runtime rather than only at review time.
Gartner projects that 40% of I&O organizations scaling AI will suffer a business-critical service disruption by 2028 without strong runtime governance. Nathan Hill, Gartner's distinguished vice president analyst and chief of research, said that "AI agents aren't just another automation script; they require a fundamentally new, proactive infrastructure governance approach."
What Good Governance Looks Like
The CDO Magazine coverage of the broader survey describes two ideas that are useful for leaders planning deployments. The first is a central authority that coordinates accountability, technology and economic decisions for AI, which Gartner's commentary calls a "Flywheel Governor." The second is what the article calls "good drag": safeguards such as data quality checks, access limits, audit trails and human review that slow risky work just enough to keep valuable projects moving at scale.
Those ideas translate into concrete steps. Each agent needs a named owner, a defined set of permissions, logs that show what it did, and a way to halt it quickly. Organizations that wait until a production incident to add these controls will find them harder to retrofit than to design in from the start.
Why It Matters
The business case for agentic AI rests on agents doing real work inside real systems. That is precisely why governance is not an afterthought. A failed chatbot answer is an embarrassment, while an agent that changes a configuration, moves money or sends data to the wrong place can cause a serious outage or loss.
The survey data suggests that many enterprises are discovering this after they have already deployed. Gartner's prediction of a 40% disruption rate by 2028 is a forecast rather than a measured outcome, and it should be read as a warning about current practice. The companies most likely to avoid it will be those that treat runtime controls as part of the product, not as paperwork added at the end.
For business leaders, the practical question is simple. Before the next agent goes live, can the organization see what it is doing, limit what it can touch, and shut it down within seconds? If the answer is no, the adoption plan is ahead of the operating model.
