About The Decision Layer

I help executives make better decisions about AI.

For sixteen years, I have worked on the problems that sit between technology and business. Data strategy, AI governance, enterprise architecture, capital allocation. I have sat in boardrooms where multi-million pound AI investments got rejected because the framing was wrong. I have watched companies spend millions on infrastructure that no one needed. I have also watched others deploy AI with such clarity of purpose that they reshaped entire operating models.

The difference between these stories is not technology. It is decision-making architecture.

What I have learned

My career started in management accounting, which gave me something most AI practitioners lack: a deep understanding of how capital allocation decisions actually get made. I moved into data analytics, then data science, and have been working in AI since 2016, across consulting, industry, and building my own startup.

Across those sixteen years, I have seen the same patterns repeat. Organisations that treat AI as a technology project managed through IT budgets and isolated pilots. Boards that reject sound investments because the proposal was capability-first instead of problem-first. Leadership teams that cannot agree on whether AI is a cost centre or a revenue driver, when it is neither: it is an operating model decision.

I have helped companies recover from these mistakes. I have also helped others avoid them entirely by getting the decision architecture right from the start.

The four layers

The Decision Layer is built around a simple observation: AI fails not because the technology is wrong, but because organisations have not addressed four interconnected layers.

People. Do you have the right talent to execute AI, or are you hiring credentials over capability? The market is flooded with half-baked experts and newcomers claiming expertise they do not have. Getting this layer wrong poisons everything downstream.

Process. There is no value in building the most sophisticated model if it never gets embedded in business processes. The companies winning are not the ones with the best models. They are the ones whose processes are AI-native. We saw this in the digital era: neobanks challenged traditional banks not with better technology, but with better process architecture. AI-native organisations will do the same to AI bolt-on organisations.

Technology. Build versus buy. ERP modernisation. CRM integration. Security architecture. AI cannot be bolted onto a legacy technology fabric and expected to deliver value. This layer examines how the technology stack must evolve for AI to be genuinely integrated, not layered on top.

Data. Data quality, evaluation datasets, model sovereignty, data ownership, proprietary data as intellectual property. Without a deliberate data strategy, AI models are built on sand. This is the layer most organisations underestimate and most vendors oversell.

What you will find here

Every piece of content on The Decision Layer respects your intelligence, embraces the real trade-offs, and helps you make better decisions when the right answer is not obvious.

I write about what I have seen work and what I have seen fail. I name real companies. I lay out the complexities. I present multiple options with their pros and cons. Then I give you my take, grounded in sixteen years of working at the intersection of data, AI, finance, and business strategy.

No hype. No pretending there are clean answers.

Making no decision is worse than making the wrong one. This is where you start making better ones.

The Decision Scientist