AI agent governance becomes boardroom risk for contact centers
LANCASHIRE, UNITED KINGDOM — Enterprise contact centers are deploying AI agents from Microsoft, SAP,Salesforce and other vendors at a pace that has outstripped basic governance — and most CX leaders cannot produce a complete list of which agents are running, what data they can access, or who authorized their permissions. The visibility gap, growing by the quarter, has now escalated into a boardroom-level risk as autonomous agents handle voice calls, resolve tickets and operate inside CRM systems with little centralized oversight.
According to a report from CX Today, the warning is being amplified by Okta CEO Todd McKinnon, who told The Verge’s Decoder podcast that AI agent governance has become a foundational enterprise security issue, not a future concern.
Why contact centers are losing track of their own AI agents
The vendor wave has arrived faster than IT and CX teams can absorb. Microsoft has embedded AI agents deep into Dynamics 365 Contact Center to autonomously handle voice and digital interactions, while SAP is automating ticket resolution at speeds no human team can supervise in real time. Salesforce CEO Marc Benioff has been defending AgentForce to investors nervous about agentic AI’s impact on SaaS, framing the technology as a platform expansion rather than a disruption.
Each vendor’s agents arrive with their own identity models, data connections and permission logic — leaving no single team with full visibility across the stack. “AI agents need to log into stuff. You need to have a system to keep track of them, define their role, define their permissions, what they can connect to and what they can do,” McKinnon said. In a contact center, those agents are routinely credentialed into CRM systems, customer data platforms and telephony layers, accessing account credentials and personal data on behalf of customers.
Building the “secure agentic enterprise”
McKinnon’s framework for governing AI agents centers on three obligations: building a complete inventory of every agent in the environment, defining and controlling connection points with precisely scoped access tokens, and applying the same discipline equally across all vendors. He argues that the right intervention when something goes wrong is not shutting down the agent itself but revoking its access — what Okta calls a “kill switch.”
“We’re pulling the access to everything the agent can access, not access to the agent,” McKinnon explained. “Almost like you would take a machine off the network.” He added a blunt warning: “Stuff is going to go wrong. There are going to be issues and threats and prompt injection.” For CX buyers, that means asking vendors directly how agent authentication is managed and what audit trail exists for every connection.
The governance gap has direct implications for the global outsourcing industry, where BPO providers running AI-augmented contact center operations on behalf of U.S. and European clients now face heightened scrutiny over how their agentic deployments are tracked, secured and revoked. Providers that build mature governance frameworks — covering inventory, access control and rapid incident response — are positioned to win enterprise trust at a moment when poorly governed AI agents are becoming a defining liability for CX buyers worldwide.
