ServiceNow CEO Bill McDermott is making a very specific argument about where AI is headed inside companies. As businesses deploy more autonomous AI agents, he says ServiceNow is becoming increasingly important. His defense is not abstract or aspirational. He is tying relevance to control, pointing to a kill switch for “rogue AI agents,” the kind that could run wild, make the wrong move, and then quietly turn an experiment into an outage.

In other words, McDermott is arguing that the enterprise AI stack is not only about capabilities. It is about governance and safety, with operational tooling that can stop misbehavior when it shows up. That is the core of the message: ServiceNow matters more because the real-world risk of autonomous agents is not theoretical anymore. Companies are not just testing chatbots. They are deploying systems that take actions, not just answers.

To understand why this matters, zoom out for a second. Enterprise software has spent years selling “workflows” and “processes.” AI has started to slip into those workflows, increasingly as agents that can trigger tasks, route tickets, update records, and respond to incidents. When AI is confined to a narrow recommendation role, failures are annoying. When AI is allowed to act autonomously across business systems, failures become disruptive. And that is where control features like a kill switch stop being a nice-to-have. They become table stakes for enterprise adoption.

McDermott’s “kill switch” framing also lands inside a boardroom reality. Boards are being asked to oversee not just growth and margin, but operational risk. Autonomous agents raise questions that traditional IT governance did not have to solve as urgently: Who is accountable when an agent makes a harmful change? How do companies detect “wrong” behavior fast enough? What is the mechanism to rapidly contain it? Enterprise platforms that can credibly answer those questions tend to be the ones that become embedded in daily operations, because they help leadership sleep at night.

There is also a regulatory angle that executives cannot ignore, even if the source does not dive into specific rules. Regulators globally are moving toward frameworks that emphasize accountability, transparency, and risk management for AI systems. Even when the details differ by jurisdiction, the direction is common: companies will be expected to manage harmful outcomes, not just claim intent. A kill switch is a concrete control that maps to that philosophy. It is the operational equivalent of a safety interlock: a way to halt something dangerous without waiting for full investigation to finish.

Second-order effects show up fast once you think about how agents get deployed. If a platform like ServiceNow can provide guardrails, then internal teams are more willing to expand agent use cases, because leadership can enforce boundaries. That changes incentives. Instead of slowing AI projects down due to risk concerns, governance tooling can accelerate them, by making the safe deployment pathway clear. Over time, that can shift budgets toward enterprise workflow platforms that integrate AI controls, not just AI models.

There is another implication for peers running similar enterprise tech stacks. McDermott is not only defending a company’s relevance. He is positioning ServiceNow as a control plane for the next phase of enterprise AI, where the ability to stop rogue agents becomes part of the product story. In a market where vendors are competing on automation, the differentiator may increasingly be how quickly systems can be constrained when automation misfires. That kind of operational credibility is hard to replicate after the fact, which means companies will prioritize it early.

For decision-makers, the strategic stake is straightforward. If your organization is deploying autonomous AI agents, you need more than dashboards and approvals. You need mechanisms that can prevent or stop harmful actions in real time. McDermott’s defense of ServiceNow’s growing importance is essentially a reminder that the enterprise AI era will reward companies that treat safety as product, not paperwork. And for executives evaluating vendor partnerships, the question is no longer just “Can it automate?” It is “Can it arrest?”