ServiceNow's Bhakti Pitre: agent kill switch needs business context

ServiceNow has spent the year pitching itself as the control tower for enterprise AI, and that ambition now extends to what happens after agents are deployed. At Okta's Oktane event, vice president of AI platform security product Bhakti Pitre framed agent containment as a governance question rather than a technical one. The company's five-step framing — discover, observe, govern, secure and measure — underpins its expanded AI Control Tower, with containment scaled to risk.

Pitre's example draws the line between an agent that stops producing results, which might need only a pause and investigation, and one that was authorized to discount only 10% but got prompt injected into discounting 100%. ServiceNow can map an agent to the business processes that depend on it, and it uses Veza's identity security technology to adjust excessive permissions. Pitre also noted that no single provider sees every layer an agent touches, from the endpoint to the network and the gateway, and said ServiceNow is working with Okta to secure agent session tokens and agent identities.

One reading of the stakes is that as enterprises move AI agents into production, containment may depend less on any one vendor's kill switch than on whether security, identity and cloud providers cooperate. Pitre's own comments suggest the technical act of stopping an agent is easy, while deciding whether to do so is where business context becomes essential — for the security teams accountable when an agent's permissions exceed what the work requires.