← back to where I focus
02 Where I focus

Responsible AI & governance

Governance isn't paperwork. Done well, it's the operating system that lets a regulated enterprise actually deploy AI — and keep deploying it as the rules, the models, and the risk surface keep moving.

Why governance is now an engineering problem

The regulatory landscape has tightened fast. The EU AI Act now classifies systems by risk and imposes binding obligations on high-risk and general-purpose AI. ISO/IEC 42001 gives organisations a certifiable management system for AI. The NIST AI Risk Management Framework has become the de facto language for talking about AI risk in board rooms. Australia, Singapore, Japan, and the UK each have their own overlays.

For a multinational, the practical question isn’t which framework to follow — it’s how to operationalise all of them at once, without grinding delivery to a halt.

The framework I build with customers

A responsible-AI programme that actually works at scale has four layers, and I build them in this order:

Where most programmes fail

I see the same failure patterns repeatedly:

The outcome to aim for

A board that can confidently sign off on AI strategy because they know the controls exist, are evidenced, and are tested. Regulators who view the organisation as a mature partner, not a target. Engineering teams who don’t fear the governance process because it’s predictable, automated, and proportionate.

Responsible AI, done properly, is what turns AI from a risk register item into a competitive advantage.