Speaking
Taking generative AI from proof-of-concept to production in regulated, financial-grade environments — before the rules were settled.
A talk for CTOs, Heads of AI and Data, enterprise architects, and engineering leaders.
Most organisations spent two years proving that generative AI works. Very few crossed the line into production. The reasons given were usually regulation, security, and uncertainty — and all three were real. But waiting was never the safe option it appeared to be. It was simply a risk that didn't show up on anyone's slide.
This talk is about what it actually takes to cross that line: how to choose a first production use case, how to sequence risk so you aren't testing everything at once, and what production teaches you that a PoC structurally cannot.
The real cost of staying in pilot, and what teams who shipped learned that the ones who waited are only discovering now.
A decision rule for picking the first thing you put in front of real users — so the unknowns you're testing are not the ones that can hurt you. Why deliberately removing your hardest risk from the first attempt is a strategy, not caution.
A PoC proves the technology works. Production proves that quality is far harder than anyone expects — especially for RAG, where the state of your source data quietly determines the quality of every answer. What this means for how you evaluate a system before you trust it.
Drawn from two years of taking generative AI use cases into production in a regulated, financial-grade European environment: retrieval-augmented generation, document intelligence, summarisation, assistants, and the responsible-AI controls around them.
Where agentic patterns break in production, and the architectural separation that keeps them traceable and auditable.
What the transparency obligations require of teams actually shipping, and where organisations believe they are covered and are not.
Responsible AI as an engineering practice rather than a review gate.
Talks are tailored to the audience. Formats: conference keynote, breakout session, executive briefing, panel, or internal leadership session.
Alireza Chegini is an enterprise AI architect based in The Hague. Over 25 years he has moved from software engineer to enterprise architect, leading large-scale initiatives in regulated, financial-grade environments where mistakes are expensive and discretion is not optional.
He designs and ships production-grade generative and agentic AI in those settings — RAG, agentic systems, AI platforms, and the responsible-AI and governance layers around them. His work is the unglamorous part of the field: what will work, what will fail, and what it will cost, established before it becomes expensive to find out.
He is also the founder of Coding As Creating, an AI-native studio building products and open, freely available AI initiatives.
Interested in having me speak?
Tell me about your audience and the slot you're programming, and I'll come back with a tailored outline.
Based in The Hague. Available across the Netherlands and Europe, in person or remote. Flexible on format.