Agentique Platform — Veolia

The context

At Veolia, AI had so far been experimented with project by project, with no shared framework. My mission: bring the group to scale by industrialising, governing and securing agentic AI for a broad, cross-functional community — developers, Product Owners, QA, UI/UX, agile project management — roughly 100 to 150 people. A scope that keeps widening month after month, with real work involved in pulling every advance into one coherent foundation.

My mission

At the core: complete AI kits, one per business function Veolia wants to advance — a dev kit, a solution-architect kit, a product kit… Each kit combines a dedicated knowledge base and custom agents, agnostic of their execution platform (GitHub Copilot, Claude or others), with default models, purpose-built tools and per-agent FinOps optimisation. A typical dev kit brings together around ten specialised agents, coordinated by an orchestrator, for a fully agentic, cost-controlled workflow.

I also work on the group’s design authority platform — standardising practices and automating scaffolding through Backstage — as well as on full LLM cost telemetry, with Azure and Google Cloud infrastructure built from scratch.

For professional confidentiality specific to this strategic project, certain details (precise architectures, data, internal mechanics) are deliberately not covered here.

What to remember

Moving agentic AI from individual experimentation to a governed, measured and adopted standard for around a hundred people: a textbook case of uniformising AI practice at large-group scale, from the business kit and the scaffolding platform down to the last cent of LLM cost.