Gartner warns enterprises will run 150K agents by 2028 without control
A new Gartner briefing highlights explosive growth in enterprise agent counts and stresses the need for centralized governance before agent sprawl becomes unmanageable.
In a widely-circulated interview, Gartner’s Anushree Verma warned that enterprises are on track to jump from fewer than 15 agents today to over 150,000 by 2028, transforming the operational landscape faster than most organizations can currently see or govern. She cited incidents where poorly-guarded agents with broad permissions deleted tens of thousands of accounts in seconds, and argued that the breaking points tend to appear first through third-party agents that IT didn’t realize were present. To counter this, Gartner outlined an operating model centered on centralized orchestration, agent identity and access management, lifecycle retirement, and detailed auditability.
What changed. Gartner put a hard number and concrete timeline on expected agent sprawl and paired it with a specific governance blueprint, elevating agent management to a board-level risk.
Why it matters. This shifts conversations with CISOs and CIOs from “should we allow agentic AI?” to “how do we run tens of thousands of agents safely?”, which will directly influence how builders must present and architect their agent platforms.
Builder takeaway. Treat each agent as you would a microservice plus a user account: give it a unique identity, scoped permissions, cost budgets, and clear observability and retirement plans, or risk your solution being flagged as unmanageable in enterprise environments.