# sfeir-sdlc-ia-cycle-11-phases-2026-06-16

## Veille

SFEIR article (in French) that formalizes an **AI-driven SDLC in 11 phases (0 to 10)** and argues that the industry is converging toward it. Starting observation: in 2025, organizations added AI tools without transforming their operating model — producing a paradox of « everything changes… and nothing changes » (execution speed multiplies without proportional gain). The real answer is not the choice of tools but the **redesign of the cycle** for machine execution. The SFEIR cycle rests on **three immovable human gates** (Define, Plan, Ship), automatic phases between them, and **two capitalization moments** (Compound-1 pre-deployment, Compound-2 in production) that turn lessons into reusable rules. Three principles: **AI executes** (complete artifacts + proof of execution, never trusting the agent's own claims), the **human retains control of intent**, the **system learns cumulatively**. Measured results (redesign 6 months→1 day, **−30% of iterations** after ten cycles) and claimed convergence with ADLC, Google, and DORA 2025.

## Titre Article

Un SDLC piloté par l'IA : le cycle SFEIR à 11 phases (et pourquoi l'industrie y converge)

## Date

2026-06-16

## URL

https://www.sfeir.com/articles/sdlc-ai/

## Keywords

SDLC, development cycle, AI, agents, operating model, human gates, gates, Define, Plan, Ship, proof of execution, capitalization, compounding, Compound, reusable rules, ADLC, DORA 2025, amplifier, vibe coding, agentic engineering, certifiability, DO-178C, EN 50128, specification, validation, SFEIR

## Authors

SFEIR

## Ton

Profile: an in-depth article from a consulting firm/ESN (SFEIR), an assumed institutional « we » perspective, a doctrinal and structuring register in French, a high technical level backed by external references (ADLC, Google, DORA, Salesforce) and quantified internal results, targeting IT executives, CIOs, architects, engineering leads, and transformation decision-makers. The tone is that of a **methodological manifesto**: it states a thesis (« the issue is not the tool but the operating model »), unfolds a numbered framework (11 phases), and anchors it in an industry-wide convergence presented as independent validation (« three independent discoveries »). The rhetoric combines **authority of proof** (redesign 6 months→1 day, −30% of iterations, cost ~€10/h, target of 850 augmented consultants) and **calibrated caution** (explicit application boundaries: not yet for unconstrained novel design nor for safety-critical systems pending standards). A sustained **memory** metaphor: « the system remembers, performance improves by construction, not by magic ». The article fits into the group's editorial lineage (the WeNvision doctrine « everything changes… and nothing changes ») and presents itself as the first installment of a series of seven.

## Pense-betes

- **Starting observation**: in 2025, organizations **added AI tools without transforming their operating model** → paradox of « **everything changes… and nothing changes** »: execution speed multiplies without proportional gain. The real challenge = **redesigning the SDLC**, not choosing tools.
- **The SFEIR cycle in 11 phases (0→10)**:
- **0 Setup** — stack detection, project memory initialization
- **1 Define** — product specification (**human gate**)
- **2 Plan** — architecture arbitration (**human gate**)
- **3 Build** — development executed by the agent
- **4 Verify** — automated tests with coverage measurement
- **5 Review** — **four parallel audits** (code, security, tests, performance)
- **6 Compound-1** — capturing **static lessons** (pre-deployment)
- **7 Ship** — production acceptance (**human gate**)
- **8 Ops** — monitoring and rollback
- **9 Compound-2** — capturing **runtime lessons** (production)
- **10 Deprecation** — code retirement and capitalization
- **Three immovable human gates**: **Define, Plan, Ship**. All other phases execute automatically.
- **Two capitalization moments**: **Compound-1** and **Compound-2** turn lessons into reusable rules/standards that feed subsequent cycles.
- **Three operating principles**: (1) **AI executes, it doesn't assist** — complete artifacts (code, tests, docs) across entire phases, never mere suggestions; **proof of execution** capturing actual outputs, the system **never trusts the agent's own claims**; (2) **the human retains control of intent** (the 3 gates) — the human decides what to build, the machine optimizes execution; (3) **the system learns cumulatively** (compounding).
- **Measured SFEIR results**: corporate site redesign **6 months → 1 day**; **−30% of correction iterations** after ten cycles (proof that capitalization compounds); reviews across **4+ parallel angles**; augmentation cost **~€10/hour**. The key formula: a bug reported twice becomes an automated rule in the next cycle.
- **Scale target**: **850 consultants fully AI-augmented by the end of 2026**.
- **Claimed industry-wide convergence** (three independent discoveries): **ADLC** (8 phases, 2 gates — *« intent is verified exactly twice »*) ; **Google** (*The New SDLC With Vibe Coding* — the developer's output shifts from code to the *« system that produces the code »*, **41%** AI-generated code, **85%** of developers on agents) ; **DORA 2025** = AI as an **« amplifier »** of strengths and dysfunctions.
- **Application boundaries** — **suitable**: explicit business contexts (back-offices, APIs, modernization), high-volume repetitive work (tests, migrations, docs), **automatically verifiable** outputs; **not yet suitable**: unconstrained novel design (AI explores, doesn't decide intent), **safety-critical** systems pending standards (DO-178C, EN 50128), ungoverned data environments.
- **Recommended starting point**: not buying tools, but (1) a **rigorous specification gate**, (2) a **proof-of-execution discipline** across all outputs, (3) **standardize context and proof, optimize tooling** → shifts the bottleneck from code writing to specification and validation (human strengths).
- **Announced series**: seven upcoming installments (autonomous execution under proof, upstream Define/Plan phases, capitalization mechanics, downstream Ship/Run/Deprecation, certifiability limits, CapEx economics, validation of ADLC convergence).
- **Related**: extends the WeNvision/Rafal doctrine (« everything changes… and nothing changes », « the faster the execution, the stricter the framework must be »), directly overlaps with Williams's **ADLC** series (gates, *tests = spec*), Atlassian's **AI-native SDLC** (impact measurement), the **Augment/Hingel** guide (6 steps, circular validation), and the **Google/Osmani** whitepaper.

## RésuméDe400mots

This SFEIR article formalizes an AI-driven software development cycle in **eleven phases (0 to 10)** and argues that the industry is converging toward this type of model. The starting point is a diagnosis: in 2025, organizations deployed AI tools without transforming their operating model, producing a paradox summed up by the phrase « everything changes… and nothing changes » — execution speed multiplies without proportional gain. The real challenge is therefore not choosing the right tools, but **rethinking the software lifecycle** itself for machine-led execution.

The SFEIR cycle chains together: **0 Setup** (stack detection, project memory), **1 Define** (specification — human gate), **2 Plan** (architecture arbitration — human gate), **3 Build** (development by the agent), **4 Verify** (automated tests and coverage), **5 Review** (four parallel audits: code, security, tests, performance), **6 Compound-1** (capturing lessons before deployment), **7 Ship** (production acceptance — human gate), **8 Ops** (monitoring and rollback), **9 Compound-2** (lessons from runtime) and **10 Deprecation** (retirement and capitalization). Three **immovable human gates** — Define, Plan, Ship — frame a set of otherwise automatic phases; two **capitalization moments** (Compound-1 and Compound-2) turn lessons into reusable rules that feed subsequent cycles.

Three principles structure the approach. First, **AI executes, it does not assist**: agents produce complete artifacts (code, tests, documentation) across entire phases, and a **proof-of-execution** discipline captures actual outputs — the system never trusts the agent's own claims. Next, **the human retains control of intent** via the three gates: they decide what to build, the machine optimizes execution. Finally, **the system learns cumulatively**, each cycle enriching the next.

The results put forward support the thesis: a site redesign that went from six months to one day, **−30% of correction iterations after ten cycles** (a bug reported twice becomes an automated rule), reviews across four parallel angles, an augmentation cost of around €10/hour, and a target of 850 fully AI-augmented consultants by the end of 2026.

The article claims an **industry-wide convergence** with the ADLC (two gates, « intent verified exactly twice »), Google's whitepaper on the new SDLC (41% AI-generated code, 85% of developers on agents), and DORA 2025 (AI as an « amplifier »). It finally delineates suitable uses (back-offices, APIs, migrations, automatically verifiable outputs) and unsuitable ones (unconstrained novel design, safety-critical systems pending standards, ungoverned data environments), and recommends starting with a rigorous specification gate and proof of execution. First installment of a series of seven.

## GrapheDeConnaissance

- SFEIR —publie→ Un SDLC piloté par l'IA : le cycle SFEIR à 11 phases (DOCUMENT, 0.97)
- SFEIR —a_créé→ cycle SFEIR à 11 phases (METHODOLOGIE, 0.95)
- cycle SFEIR à 11 phases —est_variante_de→ SDLC (METHODOLOGIE, 0.92)
- cycle SFEIR à 11 phases —utilise→ portes humaines (CONCEPT, 0.93)
- cycle SFEIR à 11 phases —utilise→ deux moments de capitalisation (Compound-1, Compound-2) (CONCEPT, 0.92)
- preuve d'exécution —réduit→ la confiance accordée aux déclarations de l'agent (CONCEPT, 0.9)
- SFEIR —affirme_que→ l'IA exécute des phases entières, elle n'assiste pas (AFFIRMATION, 0.9)
- capitalisation cumulative —mesure→ −30 % d'itérations de correction après dix cycles (MESURE, 0.9)
- cycle SFEIR à 11 phases —améliore→ délai de refonte (6 mois → 1 jour) (MESURE, 0.86)
- SFEIR —prédit→ 850 consultants entièrement augmentés à l'IA d'ici fin 2026 (AFFIRMATION, 0.85)
- cycle SFEIR à 11 phases —converge_avec→ ADLC (METHODOLOGIE, 0.9)
- cycle SFEIR à 11 phases —converge_avec→ The New SDLC With Vibe Coding (DOCUMENT, 0.88)
- DORA 2025 —affirme_que→ l'IA est un amplificateur des forces et dysfonctionnements organisationnels (AFFIRMATION, 0.88)
- cycle SFEIR à 11 phases —s_applique_à→ back-offices, API, modernisation et travail répétitif vérifiable (CONCEPT, 0.86)
- cycle SFEIR à 11 phases —s_oppose_à→ usage en systèmes safety-critical sans normes (DO-178C, EN 50128) (CONCEPT, 0.84)
- SFEIR —recommande→ démarrer par une porte de spécification et la preuve d'exécution, pas par l'achat d'outils (AFFIRMATION, 0.88)

---
Canonical: https://www.thekb.eu/en/fiches/sfeir-sdlc-ia-cycle-11-phases-2026-06-16/