AI Agent Swarms: A New Corporate Minefield
The rapid increase in enterprise AI agents by 2028 necessitates robust governance to mitigate risks and maximize value, as current controls are insufficient.
Jason Kwon ·
Enterprises are projected to manage over 150,000 AI agents by 2028, a significant increase from current numbers. This proliferation introduces substantial governance challenges and potential operational chaos.
Uncontrolled AI agent deployment, termed "agent sprawl," exposes organizations to risks including misinformation, data loss, and increased IT complexity. Current governance structures are largely insufficient, with only a small percentage of organizations reporting adequate controls.
Limiting access to AI tools does not constitute effective governance and can hinder value realization. Organizations that implemented broader access under strong governance frameworks reported significantly higher returns from their generative AI investments.
Effective governance models involve a two-tiered structure: a centralized committee for strategy and policy, and operational teams for domain-specific control implementation. Investing in third-party governance tools has also shown to nearly double the reported value from AI deployments.
Establishing clear policies for agent creation, access, and data usage is critical. A centralized inventory of all agents, coupled with continuous monitoring and adaptive controls based on risk levels, is recommended to manage this expansion.