I don't build the model. I build what has to exist underneath it for the model to ever matter — the data, the governance, the operating model that let ambitious AI and data programs actually reach production.
I've watched a lot of ambitious programs run out of boardroom patience before they ran out of ambition. Not because the vision on the slide was wrong. Because the foundations underneath it were never built to carry what was promised on top of them — and eighteen months later, the sponsor who championed the program is fielding questions about why it hasn't shown up in the numbers yet.
I started as a programmer in 2009, one year, before an MBA sent me toward strategy and delivery. Thirteen years since then, across KPMG and Capgemini Invent, I keep meeting the same failure wearing different clothes. It showed up as ESG programs that could report a number but not defend where it came from. It's showing up now, more urgently than I've seen before, as GenAI and agentic AI.
Missing the bus is the visible fear. Building on sand is the actual risk — and it's far less talked about.
Every organization I talk to knows it cannot afford to sit this one out. Almost none of them are confident the pilot they've funded will ever turn into value. And underneath that uncertainty is usually the same structural gap: the data isn't governed well enough, the operating model wasn't built for it, and the use case that worked beautifully in the lab has nowhere real to go.
That gap — between what's promised on the AI slide and what the foundation underneath it can actually carry — is where I've spent the last several years, and where I intend to keep spending them.
AI readiness through data mesh · Global pharmaceutical major
Proof-of-concept work existed across multiple business units. What didn't exist was a route from any of it into production — use cases lived with the teams that built them, data products weren't governed consistently across domains, and there was no operating model that could take a lab pilot and turn it into something a regulated pharma business would actually run on. The AI ambition was real. The foundation to carry it into production wasn't.
I program-managed the PMO and led the AI-readiness maturity assessment across a federated data mesh transformation spanning four workstreams — data products, business use cases and applications, data governance and operations, and operating model design. My role wasn't to build the AI use cases; it was to diagnose exactly where they were structurally blocked, and then direct the digital landscape and operating model overhaul that removed those blocks. I tracked the client's AI-readiness posture at each stage and used it to steer where the program invested next.
Four key AI-enabled business use cases moved into production twelve months faster than they would have on the client's prior trajectory, with a sharp reduction in the technical debt that a fragmented data and digital landscape had been quietly accumulating.
This is deliberately unglamorous work. Nobody puts "fixed the data governance operating model" in a press release. But it's the difference between an AI program that ships and one that quietly stops getting mentioned in the town hall eighteen months later.
The unglamorous layer everything else depends on — trusted, governed, decision-ready data.
The CFO and MD of a pharma major operating across the US, Canada, Europe, Russia, India and the Middle East needed a single view of the business. What they had instead was fragmented reporting nobody fully trusted.
I led the creation of a financial data lake on Google Cloud Platform (BigQuery, Cloud Storage, Data Fusion, Cloud Orchestrator) and an experiential reporting layer on Tableau, alongside the enterprise data management framework underneath it — governance principles, integration architecture — so the platform wasn't just a new system, it was a foundation.
A single, trusted reporting platform for the CFO and MD. More durably: this foundation became the base the client's later roadmap of business-critical data products was built on — including, eventually, its AI ambitions.
A data lake by itself is a commodity now. What mattered is that this one became something other things got built on top of, instead of one more system that quietly stopped being trusted within two years.
Finance leads for R&D and Manufacturing were producing management information by hand — Filings, NPV, Exhibits on the R&D side; Volumes, Yields, Opex across API and Formulation on the manufacturing side — reassembled manually every reporting cycle.
We ran an as-is study of the full MIS landscape before proposing anything, then traced every reported KPI back to its source — the mapping exercise that actually mattered, since it surfaced exceptions and breaks manual effort had quietly absorbed for years. Only then did we design and implement the client's central data lake, sized for the reporting they had and the data/AI products they'd want next.
160–200 man-days of annual reporting effort removed. More durably: a lineage-traceable reporting layer and a platform foundation the client's data-product roadmap was subsequently built on.
The bank's CFO office was working from a reporting landscape with no consistent data foundation underneath it — governance, quality and management practices hadn't kept pace with what the office needed to report.
I led and managed the PMO for an enterprise data management program that set up the data foundations — governance, quality, management — designed the new digital reporting landscape, and implemented the cloud data lake platform the CFO's office now runs on.
A reporting landscape the CFO's office could govern and trust going forward, rather than one it had to work around.
Sustainability, applied — the same foundations problem, in the regulatory register.
The bank needed to understand its exposure to a wave of emerging climate and ESG regulation — the Green Deal, Fit for 55, CRR, CRD/CSRD, Pillar 3 — and had no clear picture of whether its own data could support what regulators were about to ask for.
I led the analysis of how each regulation would land on the bank and its clients, assessed the bank's readiness for climate stress-test requirements specifically, and built a preparedness plan from a data standpoint — not a compliance checklist, a foundations assessment.
A transformation roadmap grounded in what the bank's data could actually support, rather than a paper exercise assuming the data would cooperate.
Regulatory readiness looks like paperwork from the outside. From the inside it's the identical foundations problem as everywhere else here: can the organization actually produce and defend the number a regulator, or a boardroom, is about to ask for.
In 2021, "data-driven ESG performance" wasn't a practice at Capgemini Invent — it was an idea I helped stand up from a founding team, with the organizational structure, operating model and value proposition still to be defined. Two years later it had grown from 12 people to 30, extended across Europe, North America, APAC and Australia, and become one of Capgemini Invent's recognized global offers.
I now co-lead that offer globally, and — since moving to North America — I've extended into the Data Driven Transformation practice more broadly, with my current focus on AI readiness, strategy and foundations.
None of this happens without an ecosystem:
Rarer exposure than most people at this level get, all in a supporting capacity: