Forty percent of organizations plan to demote or shut down AI agents due to governance failures, Gartner warns.

What happened: Gartner predicts that 40 percent of organizations will demote or decommission AI agents as they struggle to set up proper governance systems. The research firm says the problem stems from treating AI agent governance as binary—either locked down or fully trusted—rather than matching controls to each agent's actual level of autonomy and scope of access.

Why it matters: AI agents (systems that call on large language models to understand requests and update application data) have been heavily promoted by major software vendors including SAP, Oracle, Workday and Salesforce. However, Gartner notes that without proportional governance tailored to different trust boundaries, organizations risk either slowing down simple agents through over-restriction or exposing themselves to operational, security and compliance risk by under-restricting more autonomous agents. Vendors typically disclaim legal liability for agent actions, placing accountability on the organizations deploying them.

What to watch: Gartner recommends classifying AI agents across distinct autonomy levels, each with corresponding governance requirements including continuous monitoring, guardrails, rapid rollback mechanisms, and circuit breakers that halt operation when thresholds are breached. Organizations need clear ownership and accountability for agent behavior before scaling these deployments.