ServiceNow adds agent kill switches

Enterprise AI is hitting the awkward phase. Companies no longer have one chatbot in one sandbox. They have agents scattered across cloud platforms, SaaS apps, and internal workflows — and some of those agents can take actions, not just answer questions. ServiceNow’s update this week is basically a control-room play for that mess. At Knowledge 2026 in Las Vegas on May 5, the company expanded AI Control Tower with broader discovery, monitoring, governance, and kill-switch features for AI systems running across the enterprise. (newsroom.servicenow.com) ### What did ServiceNow actually launch? The core change is that AI Control Tower now reaches beyond ServiceNow’s own software more aggressively. ServiceNow says the product can discover, observe, govern, secure, and measure AI syste(newsroom.servicenow.com)n agent starts making bad decisions in a live system. (newsroom.servicenow.com) ### Why do kill switches matter so much? Because autonomous agents are different from ordinary software alerts. If a dashboard says a server is overloaded, a human still decides what to do. If an AI agent is allowed to approve action(newsroom.servicenow.com)ainable. (newsroom.servicenow.com) ### Where can it see now? This is the part with the most concrete weight. The expanded Control Tower adds 30 enterprise connectors spanning Amazon Web Services, Google Cloud, Microsoft Azure, plus big business software stacks like SAP, Oracle, and Workday. In plain English, ServiceNow is trying to become the place where a large company can see its AI estate in one view instead of hopping between cloud consoles and app-specific admin tools. (theregister.com) ### Why is ServiceNow pushing this now? Because the company thinks the market is shifting from building agents to governing them. Last year, at Knowledge 2025, ServiceNow introduced AI Control Tower as a centralized command center for ServiceNow and third-party AI agents, models, and workflows. This year’s move is the follow-through — less “look, we have AI” and more “here is how you keep dozens of agents from turning into operational chaos.” (newsroom.servicenow.com) ### Is this just a ServiceNow product story? Not really. It’s also an acquisition-and-platform story. Coverage around the launch ties the broader governance push to ServiceNow’s recent deals and integrations, with piec(newsroom.servicenow.com)oftware. (diginomica.com) ### What problem is it really solving? The boring answer is inventory, policy, and oversight. But that boring answer is the whole ballgame. You cannot secure or govern AI systems you cannot find. You cannot measure value from agents you cannot observe. And you definitely cannot trust autonomous workflows if there is no ob(diginomica.com)untime controls, and compliance hooks before it gets trusted at scale. (servicenow.com) ### So what’s the bottom line? This is ServiceNow trying to own the “air traffic control” layer for enterprise AI. The flashy part is the kill switch. But the bigger story is that agent sprawl is becoming a management problem, not just a model problem. If that framing sticks, the winners may be the companies that control, monitor, and shut down AI systems — not just the ones that build them. (theregister.com)