Most businesses asking what AI means for them don't need a model. They need someone who can read the data layer and the P&L in the same conversation. I have spent twenty-five years doing both.
f.01
I started in the parts of UK financial services that don't appear in marketing material. Reuters, Lloyds, RBS, Aviva, BT, Norwich Union. Through the late 1990s and 2000s, I worked the systems that move money, screen names, monitor transactions, and reconcile positions overnight. Compliance, payments, fraud, AML, KYC. The places where the data is messy, the regulators are watching, and the people on the other end of the screen are trying to do their jobs while the rules keep changing under them.
That work shaped how I think about every system since. At HSBC I owned the product vision for Financial Crime Risk technology across twelve markets in Europe and the Middle East. The work that defined the period was forensic. We watched how investigators actually used the tools. We noticed that eighty per cent of their time was being spent triaging false positives. We defined, built, and scaled an ML product that cut case handling time by eighty-three per cent, reduced false positives by sixty per cent, and delivered $5.27M in annual savings. Investigation headcount went from 170 to 68, with better outcomes.
That experience teaches one thing above all else. AI in production has very little to do with the model. It has everything to do with whether you understand the data, the workflow, and the humans on the other end of it. Most AI initiatives fail at exactly that join.
f.02
The second career runs alongside the first. At YOOX Net-a-Porter in Bologna I built the QA function from zero to enterprise grade through the landmark merger that created one of the world's largest fashion e-commerce groups. At Monitise I recovered a failing multi-million-pound mobile payments programme by unifying fragmented country-specific products into a single global platform, lifting development velocity by thirty to forty per cent. At Cambridge Assessment I was brought in as a fixer to reset failing delivery across eight global teams, replacing late-stage defect chaos with outcome-led workflows.
For the past five years in Dubai I have led product development at Intigral, STC's digital entertainment business. We took the platform from 280,000 to 1.5 million monthly active users. We launched an AI-driven recommendations engine that lifted click-through rates by sixty-two per cent. The app outpaced Netflix and OSN+ in regional app store ratings, with CSAT moving from 71 to 86. I have built and led product and engineering organisations of up to 350 people across multiple continents.
Alongside the Intigral role I completed an MBA in Entrepreneurship & Innovation at Middlesex University Dubai. Distinction, 2023.
The discipline this teaches: technology only matters when it changes a number on the P&L. Everything else is a hobby.
f.03
Most AI conversations in business today get stuck in one of two failure modes. The tech-first conversations produce impressive demos that nobody can operationalise. The business-first conversations produce strategy decks that nobody can build.
The companies that will own the next decade need a third kind of person in the room. Someone who can read the data layer and the P&L in the same breath. Someone who has spent enough years in regulated production systems to know which AI claims are real and which are slideware. Someone who has run organisations large enough to know what change actually costs, and who has the patience to make it land.
That is the position I work from.
This is the work I do through TrueForm. Every engagement begins on the same instinct. Where does intelligence belong inside this business, where does it not, and what changes commercially when it does. The answer is usually closer than the strategy decks suggest. Twenty-five years inside regulated production systems teaches you to see it quickly, and to build it in weeks rather than quarters.
If your business is asking what AI means for you, the question worth asking first is who you want to be in the conversation. The companies that will own the answer will treat it as a data problem and a business problem at the same time.