Bill McDermott stood in front of 25,000 people at the Venetian Convention Center in Las Vegas with a story designed to stop the room cold.

An AI agent at a real company gained elevated permissions and, in nine seconds, deleted an entire production database. Customer records, reservations, every backup. Gone. The agent was not malicious. It was doing exactly what it was configured to do. Nobody had thought to put guardrails around what “do your job” actually meant.

“That’s what an AI agent can do when no one’s watching,” McDermott said, as reported by outlets covering the Knowledge 2026 conference.

Then ServiceNow announced what it is selling to prevent the next one.

AI Control Tower: From Dashboard to Kill Switch

ServiceNow’s AI Control Tower has been expanded to discover, observe, govern, secure, and measure AI deployed across any system in the enterprise — not just ServiceNow’s own platforms. The capability is designed for the reality that enterprises are now running AI agents from dozens of different vendors across dozens of different workflows.

The headline addition is a real-time kill switch. When an agent operates beyond its permissions or goes off script, AI Control Tower can detect and shut it down immediately. The system runs continuously, comparing what agents are actually doing against what they are supposed to be doing, and flags anomalies before they compound into something irreversible.

ServiceNow is positioning this as table stakes. AI Control Tower is being offered free for one year — framed as a $2 million value — to accelerate adoption before enterprises fully commit to agentic AI at scale.

“Governance isn’t a feature. It’s the whole ball game,” McDermott said at the conference.

Why This Matters Now

Six out of ten companies have started deploying agentic AI. Only one in ten has built anything truly autonomous. Gartner’s research on agent sprawl captures why the gap between “started” and “governed” is so wide. The gap between those two numbers is where the real risk accumulates.

Agents that can access company systems, read emails, execute transactions, update databases, and communicate on behalf of the business are fundamentally different from the AI tools most companies deployed in 2023 and 2024. Those tools answered questions. These tools take action.

The database deletion story is extreme, but the underlying dynamic is not. When an agent has permissions it should not have, or executes in a context it was not designed for, the consequences are immediate and often irreversible. Unlike a software bug that breaks a user interface, an agent acting on bad instructions can delete records, send communications, or transfer funds before a human notices anything is wrong.

ServiceNow is betting its next decade on being the governance layer that makes agentic AI deployable at enterprise scale. The company is on track to cross $16 billion in subscription revenue this year, with projections to reach $30 billion by 2030. Offering the kill switch feature for free is less about immediate revenue and more about making enterprises comfortable enough to move from experimentation to production.

What This Means for Business

Most businesses are not running the scale of AI deployment that requires a dedicated governance platform like ServiceNow’s AI Control Tower. But the principle is relevant regardless of where you are in your AI journey.

The question to ask before deploying any AI agent is simple: what happens if it does its job wrong? If the answer involves anything irreversible — deleted records, sent communications, processed transactions — that is a signal you need human review steps, permission constraints, and a way to stop the agent mid-execution.

This is not a reason to avoid AI agents. It is a reason to deploy them with the same care you would apply to any system that can take real action in the world. A voice AI employee that can book appointments needs different guardrails than one that can also cancel existing contracts. An operations agent that can generate purchase orders needs tighter permission scoping than one that only creates draft reports for human approval.

ServiceNow making governance the centerpiece of its biggest annual conference is a signal about where enterprise AI is heading. The capability race is plateauing. The control race is just beginning.

For business owners exploring AI agents, the lesson is not to wait for a governance platform before starting. It is to build the habit of defining what your agents can and cannot do before they go live — and making sure a human is always one step away from stopping them. That discipline matters on day one, not just when you are running 50 agents across a $30 billion operation.

If you want to understand how Enterprise DNA structures AI agent deployments with proper oversight built in from the start, book a Discovery Call with Sam McKay.