Every organization knows how to manage people. There’s a reporting structure: a CEO, VPs, senior managers, managers, teams, expectations, and accountability. When something goes wrong, employees know who to report to and who owns the outcome.

But who’s responsible for your AI agents?

For most companies, the honest answer is: nobody in particular. Agents get spun up by whoever needs them for whatever task seems like a good use case. Nobody owns the portfolio.

Jon Kesman Head of Procurement - Blue Cross Blue Shield Michigan

Jon Kesman, head of enterprise procurement at Blue Cross Blue Shield of Michigan, believes that's a problem hiding in plain sight. In a recent conversation, he raised an idea that he had heard: what if organizations had someone whose job was managing AI agents the same way managers oversee people?

"That's a really interesting concept," Kesman said, "because if that stuff starts to go too unchecked inside of an organization, you're wasting a lot of time and money on things that should be better consolidated and coordinated."

It's an idea that's becoming increasingly relevant. Across organizations, employees are eager to experiment with AI, and the instinct is to adopt quickly and broadly. At Blue Cross Blue Shield of Michigan, for example, Kesman's team is working with technology partners to embed AI into processes such as reviewing statements of work and evaluating supplier contracts.

The AI Blind Spot

The risk isn't AI adoption itself—it's what happens when adoption outpaces coordination.

Kesman's example is simple. A meeting is recorded, and three different participants generate three different AI summaries.

"You've actually created more work now," he said, "because now you're expecting to read the notes to the meeting that you just sat in."

It's a small example, but it illustrates a much larger issue. When every department deploys its own AI tools without visibility into what's already available, organizations accumulate redundant agents. Over time, that translates into real but largely invisible costs.

Shadow Spend Becomes Shadow AI

For years, procurement teams watched organizations discover that sourcing and vendor management activities were happening across the business under different job titles, with little coordination or measurement. The result was shadow spend.

AI agents are shaping up to be the next version of that problem but only faster. It accumulates through well-intentioned pilots, individual experiments, and department-level decisions.

Historically, the answer hasn't been to slow innovation. It's been to assign ownership to someone who can see the entire landscape, identify redundancy, and consolidate where it makes sense before duplication becomes business as usual.

What The Research Indicates

Kesman's idea aligns with where researchers and industry analysts say enterprise AI is headed.

Industry analysts are already describing "agent sprawl" as one of the defining AI governance challenges facing enterprises. The issue isn't simply building smarter agents, but knowing which ones exist, what they're doing, and how they're being managed. A Harvard Business Review article argues that once an AI agent can take meaningful action inside enterprise systems – updating records, issuing refunds, or approving transactions – it stops being just another software tool. It introduces the same kinds of operational and accountability questions organizations already grapple with when managing people.

The emerging consensus is that AI agents need oversight: a defined identity, a clear scope of authority, and an audit trail that makes their actions explainable. That's similar to the role Kesman is describing. Yet while many organizations are developing AI governance policies, few have assigned someone responsibility for managing the growing portfolio of AI agents operating across the business.

It's Not About Slowing Deployment

Kesman sees this less as a workforce strategy than as a strategy for managing work effectively. An AI agent manager wouldn't exist to slow experimentation or deployment, but to ensure AI agents are delivering measurable value rather than quietly multiplying across the organization.

Most organizational charts don't yet include such a role. But as AI agents become a permanent layer of enterprise operations, the need for someone accountable for how they're deployed, managed, and measured is becoming harder to ignore.

The question for leaders isn't whether AI agents are already doing work inside their organization. They almost certainly are.

The real question is whether anyone can tell you how many there are, what they cost, what they do, and whether they're actually worth it.

This article is based on a conversation with Jon Kesman recorded for the Workforce Observer podcast.

Sources: Stonebranch, HBR (March 2026, Telang/Hydari/Iqbal), SHRM's State of AI in HR 2026

https://www.linkedin.com/in/jon-kesman-9b82943/