AI sprawl is coming fast as enterprises chase agents

Fortune 500 companies are expected to operate more than 150,000 agents by 2028, according to a Gartner analysis cited by TechRadar Pro. That would mark a 10,000-fold increase from 2025, when enterprises averaged just 15 agents.

That scale-up is driving governance to the top of the agenda. The piece, published as part of TechRadar Pro Perspectives, argues that while nearly every global enterprise is already using AI agents, only 12% have put a centralized governance approach in place. For most organizations, governance remains fragmented — and that fragmentation can make the problem worse.

As companies deploy a growing mix of customer-built and pre-built agents, the risk of AI sprawl rises quickly. TechRadar Pro says enthusiastic employees are increasingly building their own agents, leading to siloed use across teams and inconsistent standards. Research cited in the article says 94% of businesses are seeing more complexity, technical debt, and security risk as a result.

The article points to human-in-the-loop controls as one common response, with 52% of businesses using that model for key decisions. But it argues that approach can break down at scale, especially when different teams, regions, or departments apply their own rules for access, security, and usage.

Instead, the author argues that governance needs to be built into agentic systems from the start, not bolted on later. That means:

Recommended reading

Shadow AI grows when workplace tools get in the way

• centralized visibility into which agents are running
• insight into how they connect
• awareness of their dependencies
• a neutral orchestration layer tied to enterprise architecture and operational rules

The core idea is that manual governance cannot keep up with the speed at which agents interact with data and with one another. A centralized governance layer, the article says, would let all agents follow the same rules, use shared knowledge, and evolve with the business rather than adding to fragmentation.

TechRadar Pro frames the challenge less as a model problem than an integration one: connecting AI systems to where real work happens inside the enterprise. With adoption accelerating, the piece argues, governance has to scale with it — without slowing deployment.