AI Agent Growth Outpaces Governance Capabilities

IT leaders are experiencing firsthand what the future of AI governance looks like—and it’s not promising. A recent IBM Institute for Business Value survey found that 3/4 of IT decision-makers feel responsible for AI systems they don’t fully control.

This disconnect stems from employees and individual departments independently building and deploying new AI agents, often without centralized oversight. The survey revealed that 70% of respondents believe technology is being adopted faster than their IT teams can track it—a trend expected to accelerate with a 38% increase in organizational AI agents by next year.

“This reflects growing concerns about AI governance,” explained Matt Lieto, IBM’s CTO for Technology Innovation. “As more employees create solutions outside traditional IT channels, organizations struggle to manage the resulting assets, security risks, and business value.” While empowering employees to innovate is valuable, experts warn that unchecked experimentation can create significant challenges.

The Speed Challenge

The core issue isn’t AI adoption itself, but rather its pace exceeding governance capabilities. Ben Shneiderman, Chief AI & Analytics Officer at Domo, noted: “AI is being deployed faster than organizations can adapt their management frameworks.” He pointed out that while most implementations are well-intentioned efforts to improve efficiency, IT leaders often remain unaware of what’s happening under the surface.

For example, marketing teams might integrate LLMs into content creation, finance departments use ChatGPT for data analysis, and product teams grant AI access to customer datasets—all without CIO visibility. This creates a fragmented landscape where multiple initiatives occur simultaneously without centralized oversight.

Prioritizing Visibility & Control

Experts agree that AI governance and visibility should be top priorities for IT leaders. Shneiderman emphasized the need to “see what AI is actually doing” and manage agents like any other critical infrastructure—tracking usage, data access, costs, and issues in real-time.

“Governance isn’t optional; it’s foundational,” he said. “CIOs who succeed in the next 24 months will be those who embed visibility and policy enforcement directly into their data layers—not as an afterthought.” Similarly, Applause CTO Artisi Salvi noted that many organizations lack proper evaluation frameworks for the AI systems they deploy.

The result is often a situation where “people are producing lower-quality outputs at higher speeds,” with potential security and compliance risks amplified by decentralized control. Mind co-founder & CTO Itai Schwartz added that some AI implementations operate entirely without oversight, creating blind spots that can expose organizations to significant vulnerabilities.