SFEIR
SFEIR — Organization. claimed_capacity: equivalent to 10 000 people · headcount: ~1 000 people · internal organization: Switched to Skill Based Organisation (February 2026), up to 13 skills identified per role · position: The SDLC is the foundation, not a formality: security controls, productivity gains, the token FinOps policy, and measurement all assume a cycle with named stages · positioning: "Amplify AI rather than endure it"; adoption as a lever vs a pile of POCs · posture: "AI Only" firm, Google Cloud Premier AND Anthropic partner; thesis: the model is a commodity, value lies in Context/Harness Engineering · relation: Parent group of WEnvision · role: Author of the summary (ESN / tech consultancy) · sector: IT services company / software engineering firm (SFEIR group)
Roughly 1,000 people for a claimed production capacity of 10,000: SFEIR puts its own arithmetic on the table before anything else. The engineering firm describes itself as an "AI Only" practice, holds partnerships with both Google Cloud and Anthropic, and is the parent group of WEnvision. Didier Girard leads it. Rosalie Zandona, VP People & Culture, drove the switch to a Skill Based Organisation in February 2026, with up to 13 operational skills identified per role, on the reasoning that a frozen job description stops working as the unit of organization once tasks circulate between trades.
SFEIR publishes more than it is written about: IA et emploi : le vrai risque, c'est le décrochage, L'IA fait tomber les murs entre les métiers, Un SDLC piloté par l'IA : le cycle SFEIR à 11 phases, a rapport de recherche kill switch. One thesis runs through them: the model is a commodity, and the durable advantage sits in the engineering around it, Context Engineering, harness design, cost governance. A second says the SDLC is load-bearing. An organization without named stages has nowhere to hang its security controls, its token FinOps policy, or its measurements.
In the 11-phase cycle, Review is deliberately not a human gate. Review instruments, Ship decides, on a body of evidence rather than on the full diff. The claimed payoff, 30% fewer correction iterations after ten cycles, is a first-party internal measurement whose protocol is unpublished. SFEIR applies the same caution outward, refusing to transpose Anthropic's figures onto an IT department carrying twenty years of legacy systems.
- Type
- Organization
- claimed_capacity
- equivalent to 10 000 people
- headcount
- ~1 000 people
- internal organization
- Switched to Skill Based Organisation (February 2026), up to 13 skills identified per role
- position
- The SDLC is the foundation, not a formality: security controls, productivity gains, the token FinOps policy, and measurement all assume a cycle with named stages
- positioning
- "Amplify AI rather than endure it"; adoption as a lever vs a pile of POCs
- posture
- "AI Only" firm, Google Cloud Premier AND Anthropic partner; thesis: the model is a commodity, value lies in Context/Harness Engineering
- relation
- Parent group of WEnvision
- role
- Author of the summary (ESN / tech consultancy)
- sector
- IT services company / software engineering firm (SFEIR group)
- relations
- 68
- Cited in
- 10 fiches
Adoption measures
Neighborhood
← leads
← works at
→ publishes
→ refines
→ supports
→ created
→ draws on
→ uses
→ opposes
→ collaborates with