← All work  ·  Case 02 of 08 · POC · Completed

Hospitals track thousands of regulations. Missing one is an existential risk.

We built a continuous monitoring engine that uses HTML scraping and Claude 3.5 Sonnet to translate regulatory changes into a strict Obligation → Policy → Action workflow, drastically reducing manual review time.

Context
Enterprise Healthcare GRC SaaSAI Compliance ToolingProduct Owner2026 · Two solo POCs
Impact · Regulatory Intelligence Engine
< 1 min
Scan per source
vs. 2 days manually
Claude 3.5
Sonnet on Bedrock
For drafting legal impacts
5
Agents, monitoring pipeline
Hashed
HTML Snapshots
For precise delta detection
Scale
Scraper Bottleneck
Nested DOMs require manual setup
Chapter 01Problem space

Regulatory content is expensive to track, and missing an update is existential.

Hospitals have to track thousands of pages of federal and state regulations - from the eCFR, CMS, OIG, state medical boards, and other regulatory bodies. Before we built this engine, the process was entirely manual. Someone read through bulletins taking up to 2 days per source, trying to determine what applies to each customer and identifying what changed.

It's slow, expensive, and critical updates are easily missed.

Chapter 02Architecture

The Obligation → Policy → Action Workflow

We built a system that automatically monitors these government sources, determines what is relevant to each customer, and drafts a clear summary of what changed using Claude 3.5 Sonnet on AWS Bedrock. The system translates the legalese into a strict data model:

  • Obligation: The legal requirement the provider must follow.
  • Policy: The provider's internal document that explains how they will comply.
  • Action: The specific task assigned to fix or update a policy.

The Delta Engine: We integrated live APIs where available, but many government sites required custom HTML scrapers. When a page is scanned, it creates a hashed snapshot. On the next scan, it compares the hash against the previous version, isolates the exact HTML delta, and flags it for the LLM to process.

Gov Sources

eCFR, CMS, APIs

Monitor & Hash
Claude 3.5 Sonnet

Delta Extraction

Structured JSON
Obligation Model

Policy & Action

HITL Dashboard

Approve / Reject

Publish

Client Alert

The Human-in-the-Loop Rule: The LLM drafts the policy and impact, but it never publishes automatically. It routes to a human reviewer who sees a dashboard with the AI's confidence score and must explicitly click Approve, Reject, or Approve with edits.

Chapter 03Outcome & Caveats

Scanning time dropped from days to seconds, but scrapers don't scale infinitely.

The POC successfully reduced the time required to scan and process a regulatory update from 2 days per source to under 1 minute.

The Success

The combination of Hashed HTML snapshots and Sonnet 3.5 proved incredibly effective at pinpointing exact legal changes without hallucination.

The Bottleneck (Honest Take)

While live APIs work flawlessly, the HTML scrapers require massive manual effort to set up for deeply nested, multi-page government websites. Feeding the engine URL by URL is the current blocker to full GA scale.

Next case

Phoenix: Legacy Modernization

A solo rebuild of a 20-year-old desktop platform.

→

Let's build what actually ships.