Regulatory Change Propagation Agent

Regulated organizations routinely absorb policy changes through a fragmented manual workflow in which analysts detect updates, interpret downstream consequences, assign owners, and write change notes under time pressure. This paper proposes a Regulatory Change Propagation Agent, a multi-agent architecture that detects regulatory differences and orchestrates structured follow-up work while preserving human approval gates. The design combines a diff engine, a dependency-graph reasoning layer, and notifier-authoring agents that transform upstream legal or policy changes into targeted work packets. To evaluate the approach, the paper defines four operational metrics: detection latency, coverage of affected sections, precision of impact mapping, and reviewer acceptance rate. A prototype evaluation on a synthetic but policy-shaped corpus suggests that agentic orchestration can materially reduce time-to-detection and improve consistency of impact analysis relative to manual baselines. The contribution of the paper is not only an automation pipeline, but also a governance pattern: automation performs first-pass propagation, while human reviewers remain accountable for approval, override, and publication.