Compass
An AI native compliance workflow that blends deterministic rules with LLM reasoning, and persists an auditable decision log with a clear human approval boundary.
Problem
Compliance reviews are slow, inconsistent, and hard to audit.
AI suggestions without accountability can create risk, not reduce it.
Teams need a workflow, not just a chatbot.
Solution
A workflow that classifies content risk, suggests compliant rewrites, and attaches required disclosures.
A deterministic rules layer for predictable checks, paired with LLM reasoning where judgement is needed.
An auditable decision log and an explicit human approval gate for high risk content.
Key Engineering Decisions
Designed an explicit accountability boundary instead of pretending AI is the approver.
Persisted structured decisions (inputs, outputs, rationale) for auditability and repeatability.
Kept the system modular so rules, models, and UI can evolve independently.
What It Demonstrates
AI assisted workflow redesign, not just AI integration.
Risk based gating and traceability that maps to real compliance needs.