Abstract
This paper presents a reusable multi-agent governance measurement architecture that converts governance prose into versioned, cross-framework evidence artifacts. The architecture comprises a Standards Monitoring Agent, Control Generation Agent, and Evidence Collection Agent with mandatory human-in-the-loop review at each stage. A proof-of-concept prototype across representative NIST AI RMF and CMMC controls showed feasibility and an observed reduction of approximately 60% in manual compliance preparation effort compared to spreadsheet-based workflows.
Authors
- Jaya Kandaswamy — President, IAIMS
- Sowmya Podila — Technical Advisor, IAIMS
- Ramesh Ramaraj — Data Standards and Governance, IAIMS
Acknowledgment
The authors thank Shaadana Rajesh for her assistance with research, preparation, and presentation at the conference. The authors used generative AI tools to assist with language clarity and figure preparation. ChatGPT (OpenAI) was used for limited language refinement and editing. Gemini (Google) generated the architecture diagram from the authors' conceptual sketches. All research design, technical content, analysis, interpretations, and conclusions were developed and validated solely by the authors.