Agent sprawl: why governing AI agents defines the value of AI in the enterprise

Key takeaways

• Agent sprawl is the uncontrolled multiplication of AI agents; Gartner projects more than 150,000 per company by 2028.
• The value of AI is shifting from creating agents to governing them: by 2030, management platforms are expected to account for 80% of agent-to-agent interactions and 60% of AI value.
• An agent management platform needs observability, governance, identity and access, and horizontal orchestration across systems.
• CITSmart Hyper Agents delivers this layer: orchestration, agent chaining, multi-LLM, RAG, governance and audit trail.

What agent sprawl is

Agent sprawl is the uncontrolled multiplication of AI agents inside a company. Each area creates its own agents, in different tools, with no standard, no visibility and no governance. The result is an army of agents that no one can monitor, audit or coordinate.

The scale of the problem is what is alarming. According to Gartner, in the research Beyond Agent Sprawl: The Rise of AI Agent Management Platforms, an average Fortune 500 company is expected to go from fewer than 15 agents in 2025 to more than 150,000 agents in use by 2028. Without a management layer, this explosion creates IT complexity, risk and lost business impact.

The value of AI is shifting from creating agents to governing them

The first cycle of generative AI was about creating: models, copilots, agents. But models and agents are quickly becoming commodities. Gartner projects that, by 2030, agent management platforms will account for 80% of successful agent-to-agent interactions and capture more than 60% of the compound value of AI. The competitive advantage no longer lies with whoever creates the agent, but with whoever controls execution.

In practice, traditional automation, which only orchestrates workflows, gives way to platforms that coordinate intelligence: they generate the workflows, speed up decisions and apply governance so that autonomous work stays transparent and aligned with the business.

What an agent management platform needs to have

Gartner describes the agent management platform (AMP) as the central hub where governance, performance and value meet. It brings together six elements: security, ready-made libraries, tools, a control panel, a marketplace and agent observability.

Observability is the most critical part: testing new agents, managing the lifecycle, monitoring performance and maintaining audit trails. It is what ensures agents are reliable, effective and auditable, from the first deployment to retirement.

• Horizontal orchestration: coordinating agents that live in different systems (ERP, CRM, ITSM) to complete a task end to end.
• Governance and audit trail: a trust center that enforces policy, security and compliance, and turns telemetry into auditable evidence.
• Identity and access: each agent with its own identity and specific permissions, secure delegation of actions and traceable access revocation.
• Observability and lifecycle: testing, performance monitoring and continuous auditing of each agent.

Trust, risk and identity: the brakes that unlock scale

Paradoxically, control is what makes scale possible. Gartner is direct: what holds back the adoption of agents at scale are concerns about trust, risk, security, identity and governance. Agents that act with elevated permissions and little human oversight widen the organization's attack surface.

That is why each agent needs its own identity and a clear scope of permissions. Sensitive actions require controlled delegation and, when necessary, human approval. And everything needs to be recorded, so that autonomous execution is auditable and aligned with business rules and regulation.

How Central IT solves agent sprawl with CITSmart Hyper Agents

This management and governance layer is exactly what CITSmart Hyper Agents delivers. Instead of yet another tool that creates isolated agents, Hyper Agents is the platform for creating, connecting, orchestrating and governing AI agents on top of the CITSmart platform, in a low-code environment.

• Orchestration and agent chaining: the output of one agent feeds the next, forming execution pipelines that cross systems, instead of loose agents per area.
• Multi-LLM and RAG: each agent uses the right model and answers with the context of corporate data, not guesswork.
• Native governance: role-based access control, sensitive actions with human approval and an end-to-end audit trail, the trust center Gartner describes.
• Scope and tools per agent: each agent only accesses the tools it is given, which keeps identity and permissions under control.

For critical security and infrastructure operations, CITSmart Autonomous AI puts these agents to work, with governed execution and a human in the loop when it matters. And because everything runs on the CITSmart platform, the agents connect to the same ecosystem that already manages services (ITSM and ESM), public administration (GRP) and business processes. This is the horizontal orchestration that prevents islands of automation.

Conclusion: managing agents is the new competitive advantage

The AI race is changing focus. The winner is not whoever has the most agents, but whoever can manage, govern and coordinate them at scale. Agent sprawl is the price of adopting AI without a management platform; governance is what turns scattered agents into a reliable digital workforce. With CITSmart Hyper Agents, Central IT puts this control layer at the center, so that AI produces auditable results, not chaos.