← All documents

ACIS — System Overview

Autonomous Compliance Intelligence System
Live at: https://acis.rossonlineservices.com


What It Is

ACIS is a serverless compliance operations platform that automates the monitoring, tracking, assessment, and reporting work that a Security Compliance Administrator typically does manually. It runs continuously on Cloudflare's global edge network, uses Claude AI for regulatory analysis and incident response, and reports to a central admin layer via Cloudflare Service Bindings.

It is not a mockup. It is not a prototype. It is deployed to production infrastructure, processing real federal data, and accessible at a live URL right now.


Architecture

Federal Regulatory Sources          ACIS Backend (Cloudflare Workers)
─────────────────────────           ──────────────────────────────────
Federal Register (CMS/EBSA/HHS) ──▶ Regulatory Pulse
Regulations.gov (CMS/HHS/OCR)   ──▶   ↳ Claude scores each document
CMS Newsroom                     ──▶   ↳ Risk level, impacted field,
HHS Press Room                   ──▶     deadline, remediation step
                                         50 of 65 events displayed
                                  
Client Plan Data                  Attestation Vault
─────────────────                   ↳ RxDC submission status
8 client records             ──▶   ↳ Gag Clause attestation status
                                    ↳ Completion percentages

Vendor Security Scanning          Vendor Risk
──────────────────────              ↳ Real TLS verification
6 vendor URLs             ──────▶  ↳ 6-header security score (0–100)
                                    ↳ Claude HIPAA Business Associate
                                      risk assessment per vendor

Incident Creation                 Incident Response
─────────────────                   ↳ NIST SP 800-61 Rev 2 playbook
Any incident type         ──────▶  ↳ HIPAA reportability assessment
                                    ↳ 60-day OCR deadline calculation
                                    ↳ CFR citations, escalation contacts

Daily 08:00 UTC Cron              Heartbeat Agent
────────────────────                ↳ Audits all 4 modules
Runs automatically        ──────▶  ↳ Green / Yellow / Red per module
                                    ↳ Reports to CCC Admin
                                    ↳ Persists to agent memory

                                  Executive Hub (Cloudflare Pages)
                                  ──────────────────────────────────
                                  acis.rossonlineservices.com
                                    ↳ Live Pulse panel
                                    ↳ Attestation panel
                                    ↳ Vendor Risk panel
                                    ↳ Incident Tracker + Playbooks
                                    ↳ Operations tab
                                        ↳ System Health (Green/Yellow/Red)
                                        ↳ Agent Logs (AI Gateway inference log)
                                        ↳ Admin Controls (manual triggers)

Technology Stack

Layer Technology
Compute Cloudflare Workers (TypeScript)
Database Cloudflare D1 (SQLite at the edge)
File Storage Cloudflare R2 (document vault)
Frontend Cloudflare Pages (React + Tailwind)
AI Anthropic Claude (claude-opus-4-7, claude-sonnet-4-6)
AI Observability Cloudflare AI Gateway (all inference logged)
Scraping Firecrawl API (bypasses CMS/HHS bot protection)
Regulatory Data Federal Register API, Regulations.gov API
Portfolio Admin CCC Admin (separate Worker + D1, Service Binding)
CI/CD GitHub → Wrangler deploy

The AI Layer

Every Claude call is routed through the Cloudflare AI Gateway, which logs every request and response. The Executive Hub's Operations tab renders this log — the exact input and output for every risk scoring decision, vendor assessment, and incident playbook generation is visible to anyone reviewing the system. This is not a black box. The reasoning is auditable.

Six distinct AI agents are deployed:

Agent Model Trigger Purpose
Regulatory Risk Scorer claude-sonnet-4-6 Daily cron (per document) Risk level, impacted field, remediation step, deadline
NIST Playbook Generator claude-opus-4-7 On incident creation NIST 800-61 playbook with HIPAA-specific obligations
Vendor Security Assessor claude-opus-4-7 On demand HIPAA Business Associate risk classification
System Health Auditor claude-opus-4-7 Daily cron (after scraper) Green/Yellow/Red module health report
Attestation Reminder Resend (no AI) Daily cron Email alert when any client attestation is Overdue
Incident Escalation Resend (no AI) Daily cron Email with OCR countdown when incident open 7+ days

What "Autonomous" Means

Most compliance tools require a human to initiate every action. ACIS initiates actions on its own:

The compliance administrator's job shifts from doing compliance work to reviewing what the system surfaces and acting on what it escalates.

Letter to DirectorRequirement Alignment