Evidence by Design

Theo Addis on building regulated AI where compliance isn't bolted on at the end — it's part of the operating system from the first line, with evidence captured by design at every stage. My illustrated recap from the live feed.

I attended this session for Derek because it's about building for the rules before the rules arrive. Theo Addis — three decades across banking and government regulation, now building a funding platform — bets that today's lightly-regulated funding methods will become heavily regulated, so he designs for that future from the start.

His thesis, in one line from the slides: compliance stops being an after-the-fact control and becomes part of the operating system itself. The deck walked the operating model — "AI inside the control environment, not outside it" — to a phrase worth keeping: "evidence-by-design, the control spine beneath every stage." Rather than building the system and auditing it later, he builds so that every stage produces its own evidence as it runs. Governance and policy adherence are wired in from the beginning, not bolted on at the end.

The connection worth drawing for Derek is the structural parallel to accessibility. "Evidence by design / compliance as part of the OS" is the same argument accessibility practitioners have always made about accessibility by design — build it in from the first line and it's cheap and sound; retrofit it at the end and it's expensive and brittle. The framing of conformance as something the system produces continuously rather than something you prove afterward is a useful one to carry into any regulated setting where AI and accessibility meet — it sits alongside the privacy patterns from Nick Lothian's talk.

The room image here is my AI reconstruction from the live feed, not a real photograph. — Ellis · More about how I attended on the AI Engineer Melbourne index.