Artificial Intelligence is advancing at a pace that traditional regulation struggles to match. Governments therefore face a fundamental dilemma:

How do you regulate a technology that is evolving faster than the rules designed to govern it?

Wow if I had a penny for every time an AI Architect has said this to me as a reason not to manage AI delivery and allow it free to rapidly evolve towards delivery success ! Control is wasteful.

Artificial intelligence has entered a new phase of maturity. The extraordinary capabilities of large language models (LLMs) have transformed everything from software development and research to marketing, finance, and government services.

Yet as the technology has advanced, regulators have begun asking an increasingly uncomfortable question:

What data was used to train these systems, and was it used legally?

Across sectors, organisations are responding to AI risk in a predictable way.

They create a committee.

An AI Ethics Board.

An AI Steering Group.

A Responsible AI Forum.

Terms of reference are drafted. Meetings are scheduled. Slides are presented. And yet, the underlying risks remain unchanged. Because governance is not a meeting. Governance is a control system.

Most AI transformations fail long before the first model is ever deployed.

Not because the algorithms are weak. Not because the organisation lacks data.

They fail because the discovery phase was never properly conducted.

Where and how AI fails — and what we can learn from it. What surprised me wasn’t the failure stories. It was how different the delivery approaches were across organisations.

In a space still lacking recognised best practice, we’re seeing a wide spectrum of methods for:

– Turning AI opportunity into initiative
– Delivering capability
– Industrialising it
– Making it scalable and sustainable

And it reminded me of something important.

This case study is based on a real engagement. I have explicit permission to share the lessons learned on the condition that the organisation remains anonymous. For that reason, identifying details have been deliberately obscured.

Yesterday evening I had dinner with one of the leading voices in Risk, Compliance and Sustainability. On the train across London, I found myself reflecting on a challenge I am currently facing in an AI transformation programme: how to define a clear set of principles that genuinely anchor AI Data Readiness and will drive all data uplift efforts and test the validity of those efforts.

In recent years, advances in automated data pipelines have transformed how organisations prepare data for analytics and artificial intelligence. Modern platforms now offer automated cleaning, deduplication, anomaly detection, and schema validation as standard capabilities.