Project Overview
This project focuses on continuously monitoring healthcare regulatory changes (HHS and CMS), detecting what has changed over time, and delivering concise, role-specific summaries to stakeholders.
Problem Statement
Healthcare organizations live under a constant stream of regulatory change from agencies like HHS and CMS. New rules and updates arrive in the Federal Register every week, often as dense, technical documents.
Today, many compliance, billing, and IT teams track this manually: subscribing to alerts, skimming PDFs, forwarding emails, and hoping someone notices the one rule that actually affects their workflows, billing logic, or data handling.
Why This Creates Real Risk
- Missed CMS payment rules can imply billing errors, audits, or clawbacks.
- Missed HHS privacy and security changes can lead to potential HIPAA issues, investigations, and reputational damage.
- Late awareness for engineering and product teams means changes surface when it is more expensive to adapt systems.
How can we automatically monitor HHS and CMS rule changes, compare them across time, and turn them into concise, role-specific insights that show up where people already are (email), instead of forcing everyone to read the Federal Register?
It is interesting because it combines real-world regulation, live external data, multi-step reasoning, and the need to tailor outputs to different stakeholders in the same organization.
Why Agents
A simple single-call LLM is not enough here. The workflow is inherently multi-step and multi-role:
- Fetch real documents from the Federal Register for specific agencies and dates.
- Filter and structure them into Final vs Proposed rules.
- Compare this period vs the previous period (for example, last 30 days vs the 30 days before that).
- Interpret the impact differently for compliance, IT, and engineering.
- Push the results out via email.
Agents are a good fit because they let me break this into composable capabilities, each with a clear responsibility:
- Data agent: specializes in talking to the Federal Register API.
- Comparator agent: specializes in analyzing trends and changes across time windows.
- Orchestrator agent: decides when to call each, and how to turn raw outputs into stakeholder-ready narrative.
Using A2A (agent-to-agent) with Google ADK lets these agents run as separate services, communicate over HTTP, and still be invoked in a natural way from the orchestrator. This mirrors how you would build a production system: specialized services talking to each other, not a single giant script doing everything.
Demo
A short walkthrough of the solution and how the agent workflow produces stakeholder-ready summaries.