Researchers propose five-level AI agent governance framework validated through 750 simulations

Researchers at arXiv have published an academic framework designed to address the rapid, uncontrolled adoption of autonomous AI agents in enterprise operations. The paper, authored by Vivek Acharya, introduces the Agentic AI Governance Maturity Model (AAGMM)—a five-level structure intended to help organizations manage what the authors call "agent sprawl," referring to redundant, ungoverned, and conflicting AI agents proliferating across business functions.

The framework spans 12 governance domains and grounds its structure in existing standards: NIST's AI Risk Management Framework and ISO/IEC 42001. The authors present a taxonomy of agent sprawl patterns—functional duplication, shadow agents, orphaned agents, permission creep, and unmonitored delegation chains—each with associated business cost implications. The model is designed to link governance capability directly to measurable organizational outcomes.

The authors conducted 750 simulation runs across five enterprise scenarios to test their framework. The simulations measured four categories of business impact: cost containment, risk incident rates, operational efficiency, and decision quality. The results report statistically significant differences (p < 0.001) with effect sizes exceeding 2.0 across all governance maturity levels. Organizations at levels 4 and 5 showed 94.3% lower sprawl indices compared to level 1, along with 96.4% fewer risk incidents and 32.6% higher effective task completion rates.

The paper cites industry survey data indicating that only 21% of enterprises currently have mature governance models for autonomous agents, and projects that 40% of agentic AI projects will fail by 2027 due to inadequate governance and risk controls. The authors position the AAGMM as an actionable roadmap for practitioners seeking to govern autonomous agents while improving business outcomes.