Enterprises Winning With AI Agents Are the Ones Reining Them In

Quick Facts

• Half of enterprises have deployed an AI agent that passed internal evaluations but still caused a customer-facing failure, per a June 2026 VB Pulse survey of 157 enterprise respondents.
• Only 13% of organizations believe they currently have the right AI agent governance in place, even as Gartner projects the average Fortune 500 company will run more than 150,000 agents by 2028.
• Gartner predicts 40% of enterprises will demote or decommission autonomous AI agents by 2027 due to governance gaps discovered only after production incidents.

The enterprises getting real results from AI agents are not the ones giving agents the most freedom. They are the ones building deliberate constraints around them. That is the central finding from VentureBeat’s research series drawing on five surveys of 573 enterprise leaders.

The data reveals a dangerous gap between deployment speed and the controls needed to manage what gets deployed. Half of enterprises in a June 2026 VB Pulse survey reported that an AI agent or large language model feature passed internal testing and still caused a customer-facing failure. One in four said it happened more than once.

Trust in Testing Is Nearly Nonexistent

Only 5% of enterprise leaders say they fully trust their automated evaluation processes. The most common complaint, cited by 29% of respondents, is that evaluations align poorly with real-world outcomes. Yet two-thirds of organizations are moving toward zero-human-in-the-loop deployment anyway, either already doing it for low-risk agents or actively building toward it within a year.

Post-deployment monitoring is equally thin. Only 23% of enterprises run real-time quality checks on agent outputs once agents are live. Another 51% monitor only system health, meaning uptime and request logs. Those metrics confirm an agent is running. They say nothing about whether its answers are correct.

Most Deployed Agents Are Not Agents

Seventy-one percent of enterprises said a quarter or fewer of their deployed agents can complete multi-step work on their own. Only 10% said true autonomous agents make up the majority of what they run. Gartner has flagged this as a widespread problem, warning that the most common misconception is calling AI assistants agents, a misunderstanding it labels agentwashing.

METR’s 2025 study found frontier agents succeed nearly 100% of the time on tasks that take humans under four minutes, but less than 10% of the time on tasks exceeding four hours. The Stanford AI Index found autonomous agent deployment across business functions still in single digits, meaning benchmark gains are running far ahead of production use.

Security Controls Are Lagging

The security picture is worse. Credential sharing persists across nearly two-thirds of agent fleets. Barely one in five organizations isolates its highest-risk agents. And 53% have already experienced a confirmed agent security event or near-miss. Deloitte’s 2025 survey found 74% of enterprises expect significant agent adoption within two years, yet only 21% say their governance is mature.

Raj Koneru, founder and CEO of Kore.ai, called the phrase we have guardrails the most dangerous sentence in enterprise AI. Qualtrics CSO Assaf Keren identified a structural problem: organizations are introducing non-deterministic decision-making into environments built for deterministic processes. Keren noted that 22% of Qualtrics’ SOC triage is now AI-driven, with no codified threshold separating what an agent can auto-execute from what requires human review.

What Winning Looks Like

Kristof Horompoly, head of AI at ValidMind, put it plainly: the fix is not better models. It is narrowing scope to bounded, well-instrumented tasks and building real evaluation processes before scaling. Vlad Luzin, CTO and co-founder of BAND, said mature deployments can reach 80 to 90% automation, but human oversight remains fundamental.

Shiva Varma, senior director analyst at Gartner, traced the root cause to binary thinking: enterprises treat agent governance as either locked down or fully trusted, and that framing drives failure. Writer data shows 36% of organizations have no formal plan for deploying agents at all, and 35% admit they could not shut down a rogue agent if one appeared.

The global AI agents market was estimated at $7.6 billion in 2025 and is projected to exceed $10.9 billion in 2026. Analysts expect the multi-agent segment to grow at a 48.5% compound annual rate through 2030. Companies building governance infrastructure now will have a structural advantage as that market scales.

Read more: Enterprises winning with AI agents are limiting how much the agents can do alone