Applied AI
AI with an owner, a data foundation and a number to beat.
Most businesses with an AI strategy have a pilot and a slide deck. Very few have a system in production that moves a line in the P&L.
I delivered three production AI systems in a GxP-regulated pharmaceutical logistics business across 26 countries. GDPR and GMP compliance were in scope throughout.
- A RAG-powered sales enablement tool that cut engineering information requests from sales by 45%.
- A pre-multimodal computer vision system for container damage assessment.
- AI network modelling that safeguards life-critical pharmaceutical shipments across 26 countries, integrated with Power BI.
- AI network modelling that safeguards life-critical pharmaceutical shipments across 26 countries.
The sequence mattered. A governed data strategy and a cloud-based big data platform came first. AI followed, because AI on ungoverned data is a demo.
Hands-on depth
Azure OpenAI, LangChain, Microsoft Copilot Studio, RAG pipelines, agentic workflows. I build with these. I do not only advise on them.
Agentic engineering at scale
AI agents went into the engineering function itself. Automated code review and test generation ran across the delivery lifecycle, cycle times came down and developer throughput went up. The same approach now runs through my advisory work.
What it covers
- Triage of AI ideas by commercial return before money is spent
- A straight assessment of whether your data can carry what you want to build
- Build, buy or configure decisions, and holding vendors to them
- Taking a system from pilot to production with governance in place
- AI in the engineering lifecycle, including code review and test generation
No AI theatre. A pilot that demos well is not a result. A result has an owner, governed data behind it and a number it had to beat. The 45% is the test.
Further reading. Before You Do AI, Build the Foundations That Will Sustain AI Success
Academic method into production. Across a three-year Knowledge Transfer Partnership with Cardiff University Business School, where I was the named business supervisor, data science methodology became a production AI network optimisation solution embedded in day-to-day operations and strategic decisions across pharmaceutical cold chain supply chains.
Start with the problem, not a job title.
Tell me what is happening, what is coming and what nobody in the business can currently answer.