77% of IT managers say their AI agents are out of control - 5 ways to rein in yours

Editor's take

A significant majority of IT managers report their AI agents are operating without adequate oversight, leading to the proliferation of unauthorized AI tools across organizations. This situation poses substantial risks regarding data security, compliance, and the efficient allocation of IT resources, as shadow AI deployments can bypass established governance frameworks and lead to unexpected costs or vulnerabilities. The sheer volume of these unsanctioned tools suggests a growing gap between the rapid adoption of AI capabilities by end-users and the IT department's ability to manage and secure them.

The implications extend beyond mere IT headaches; it highlights a fundamental challenge in the enterprise AI adoption lifecycle. As companies like Microsoft and Google push more sophisticated AI assistants into the hands of employees, the onus is increasingly on IT to understand and control these emergent applications, much like they did with the rise of SaaS. This uncontrolled growth could hinder the realization of AI's full potential by creating fragmented, unmanaged deployments that are difficult to integrate and scale effectively.

Moving forward, the focus will likely shift to developing robust AI governance platforms and clear policies that balance innovation with control. Organizations will need to monitor how effectively IT departments can identify and integrate sanctioned AI agents, perhaps through centralized marketplaces or approval workflows, to mitigate the risks associated with the current "wild west" scenario. The success of these measures will determine whether enterprise AI becomes a managed strategic asset or a persistent operational risk.