Building an AI Compliance Brain

Four hours a day. That's how much time compliance requirements eat right now. Regulatory queries, cross-referencing documents, verifying rules across EU-ETS, UK-ETS, and CORSIA frameworks. It's essential work — get it wrong and airlines face penalties — but it's also repetitive, pattern-based, and perfect for AI.

Today we're bringing a new machine online: The Counsel.

The Concept

The Counsel is a 2012 Mac Mini (£30, eBay) running Ubuntu Server with a 500GB SSD. Its job: become the definitive compliance knowledge base for aviation emissions.

The architecture:

• Vector database — all compliance documents chunked, embedded, and searchable by meaning (not just keywords)
• RAG pipeline — ask a question in plain English, get the relevant regulation with source citations
• Multi-scheme support — EU-ETS, UK-ETS, CORSIA, SAF rules, all in one queryable system
• Integration with The Auditor — cross-reference regulatory requirements against actual flight data

Why This Matters

Compliance isn't optional. Airlines must report emissions accurately, on time, to the correct scheme. The rules differ by jurisdiction — EU-ETS has proportionality calculations, UK-ETS doesn't. CORSIA covers international routes. SAF (Sustainable Aviation Fuel) has its own claiming process.

Right now, a human (me) holds most of this knowledge. That's a single point of failure. If The Counsel can answer "What are the UK-ETS reporting requirements for a domestic flight using SAF?" with the correct regulation cited — that's four hours a day returned to me and, eventually, a product feature for customers.

The Build

This post will be updated as we build. Follow along — every step, from hardware setup to first successful query.