WEnvision

Olivier Rafal, Consulting Director Strategy at WEnvision, argued on CIO-Online in February 2024 that generative AI belongs to product engineering rather than data science: building a foundation model costs months, millions of euros and enormous datasets, so most firms should assemble a platform of orchestrators and APIs and fund it as a monthly flow rather than a one-off investment.

WEnvision is a French consultancy inside SFEIR, working from Paris, Lille, Strasbourg, Bordeaux, Nantes, Toulouse, Belgium and Luxembourg. Its published output splits between market reading and field method.

Rafal handles the market reading. He treats OpenAI's ChatGPT Atlas as the start of conversational web access that employees will expect at work, and the Linux Foundation's Tokenomics Foundation as the opening of FinOps applied to AI, citing Goldman Sachs on global token use multiplying by 24 by 2030. On 1 June 2026 he described SFEIR retiring "double pizza teams" in favour of "sandwich teams": a business expert and tech lead pair that now carries roughly 80% of the production chain.

Antoine Habert supplies the field record. Two years deploying agents in a European fintech through 2025 produced 100% automation of level-1 support and four architectural pillars (reasoning transparency, organizational memory, cognitive collaboration, adaptive supervision), formalized in RAISE. His PROJ-AI method, measured across 3 missions actives, 31 décisions enregistrées, reports onboarding falling from three weeks to two days.

The firm also accompanied the Almia rollout at AG2R LA MONDIALE. Habert's own caveat bounds the whole programme: "technologie 20%, discipline d'équipe 80%".