# suchaud-wenvision-fin-valeur-faisabilite-prioriser-ia-2026-10-22

## Veille

Article by **Nicolas Suchaud** (**WEnvision**) published on **October 22, 2026** on the firm's Insights hub (cluster "AI in Enterprise & ROI", ~9 min), accompanied by **six visuals**: a header image and a carousel of five numbered slides (a summary slide, then one slide per grid). Thesis: the **Value × Feasibility** matrix no longer sorts AI projects, because almost everything can be prototyped quickly and ends up as a *quick win*.

## Titre Article

La fin de Valeur × Faisabilité : pourquoi la matrice produit ne sait plus prioriser l'IA

## Date

2026-10-22

## URL

https://www.wenvision.com/fr/articles/la-fin-de-valeur-faisabilite-pourquoi-la-matrice-produit-ne-sait-plus-prioriser-lia/

## Keywords

AI prioritization, Value × Feasibility matrix, quick wins, RICE, MoSCoW, WSJF, Value / Effort, decision portfolio, decision rights, AI Decision Matrix, ambiguity, risk, frame act learn, Adopt Partner Build Buy, differentiation, market availability, reversibility, six signals, Work Trend Index 2026, four work modes, delegation, collaboration, organizational maturity, AI committee, One New Zealand

## Authors

Nicolas Suchaud — WEnvision (cabinet de conseil) ; hub Insights, rubrique Intelligence Artificielle.

## Ton

First-person consultant post, structured for quick reading: TL;DR up front, three reasons, three grids, a summary table, four recommendations, a closing question to the reader. The title is owned as *"deliberately provocative"*: the author then clarifies that value and feasibility don't disappear but are no longer enough to decide. The text draws on three 2026 publications it cites precisely (authors, dates, survey method) and carries over their caveats: a statistical, not causal, correlation for Microsoft, qualitative placement for the four modes. The author's own contribution lies in assembling the three grids into three questions and in the signal said to be prioritized first, reversibility. The visuals, in carousel format (light background, blue titles, black box carrying one maxim per slide), are designed for distribution on social networks.

## Pense-betes

- **Observation**: *"A matrix where everything is a quick win is no longer a matrix. It's a list."* Example: Meta showed the first prototype of Muse to its board after one to two weeks.
- **Costs shifted**: production reliability, governance, adoption — invisible at the point of arbitration.
- **Same capability, two decisions**: a price recommendation on an entry-level range vs. on a three-year framework contract.
- **Microsoft WTI 2026**: 20,000 employees, ten countries; organizational factors **67%** vs. individual **32%**; **26%** see their leadership durably aligned; **19%** in the *Frontier* zone.
- **Header image**: RICE, MoSCoW, Value × Effort, and **WSJF** crossed out ("prioritize what we can build") → three cards ("prioritize what we agree to delegate"). WSJF appears only in the image.
- **Slide 2/5 (MIT)**: reproduces the original matrix with Frame / Act / Learn pictograms. Routine: AI on all three; Consequential: AI frames and acts, human learns; Exploratory: human + AI framing, AI acts, human learns; Strategic: human + AI framing and action, human learns. Maxim: *"Prioritize decisions, not use cases."*
- **Slides 3 and 4/5 (BCG)**: differentiation × availability matrix with six capabilities placed; table of the six signals; the author keeps three of them (differentiation, urgency, reversibility). Maxims: *"Build only what differentiates"*, *"Keep reversibility as long as nothing differentiates"*.
- **Slide 5/5 (Microsoft)**: axes human directs → supervises, agent assistant → teammate; delegation (directs + teammate), quick question (directs + assistant), collaboration (supervises + teammate), exploration (supervises + assistant). Maxim: *"Choose the mode before choosing the tool."*
- **Summary table**: MIT replaces *"is it feasible?"*, BCG *"how much does it cost to develop?"*, Microsoft *"how many users?"*.
- Related to [[habert-wenvision-proj-ai-repo-agent-ide-doctrine-2026-05-05]].

## RésuméDe400mots

Nicolas Suchaud, of WEnvision, starts from a tool every product team knows: the Value × Feasibility matrix and its variants (RICE, MoSCoW, Value / Effort). These methods assume that building is expensive. With generative AI and then agents, that assumption collapses: an agent prototype can be put together in a few weeks, and most use cases end up in the top-right corner. The grid no longer sorts.

Three reasons. Feasibility no longer discriminates at the prototype stage: an LLM, a few tools connected via MCP, and a chat interface are enough, and the real costs (production reliability, governance, adoption) don't show up at the point of arbitration. Value depends on context: according to MIT CISR, the same AI capability is safe in one context and risky in another. The constraint is organizational: in Microsoft's Work Trend Index 2026, culture, managers, and HR practices account for 67% of AI's reported impact versus 32% for individual factors — a correlation, not a causal effect.

The author proposes prioritizing decisions rather than use cases, and assembles three grids published in 2026. The AI Decision Matrix by Sebastian, Weill, Haskamp, and vom Brocke (MIT CISR) crosses ambiguity and risk and splits the three stages of a decision — frame, act, learn — between human and AI. It distinguishes routine decisions (automate), consequential decisions (make reliable), exploratory decisions (experiment), and strategic decisions (steer), illustrated with One New Zealand. BCG's grid crosses differentiation and market availability to choose between adopt, partner, build, or buy, supplemented by six signals and one principle: the lightest structure that provides the necessary access. The author singles out reversibility first. Microsoft's four modes (delegation, collaboration, quick question, exploration) classify ways of working; a use case's value changes depending on the mode, and only 19% of users are in the Frontier zone.

The five-slide carousel reproduces each source diagram with its caption and a maxim. The text concludes that value has broken down into three questions and sets out four recommendations: replace the use-case backlog with a portfolio of business decisions, build only what differentiates, measure the organization's maturity, choose the mode before the tool. Scarcity has shifted toward trust, differentiation, and the capacity to change the way of working.

## GrapheDeConnaissance

- Nicolas Suchaud —travaille_chez→ WEnvision (ORGANISATION, 0.95)
- Nicolas Suchaud —publie→ article La fin de Valeur × Faisabilité (DOCUMENT, 0.98)
- Nicolas Suchaud —affirme_que→ la faisabilité technique ne discrimine plus les cas d'usage IA : presque tout devient quick win (AFFIRMATION, 0.95)
- Nicolas Suchaud —s_oppose_à→ matrice Valeur × Faisabilité (METHODOLOGIE, 0.9)
- priorisation par portefeuille de décisions —remplace→ matrice Valeur × Faisabilité (METHODOLOGIE, 0.88)
- priorisation par portefeuille de décisions —utilise→ AI Decision Matrix (METHODOLOGIE, 0.92)
- priorisation par portefeuille de décisions —utilise→ grille Adopt Partner Build Buy (METHODOLOGIE, 0.92)
- priorisation par portefeuille de décisions —utilise→ quatre modes de travail avec l'IA (CONCEPT, 0.9)
- MIT CISR —a_créé→ AI Decision Matrix (METHODOLOGIE, 0.95)
- Boston Consulting Group —a_créé→ grille Adopt Partner Build Buy (METHODOLOGIE, 0.95)
- quatre modes de travail avec l'IA —fait_partie_de→ Work Trend Index 2026 (DOCUMENT, 0.94)
- Microsoft —publie→ Work Trend Index 2026 (DOCUMENT, 0.97)
- Work Trend Index 2026 —mesure→ facteurs organisationnels 67 % vs individuels 32 % de l'impact déclaré de l'IA (20 000 salariés, 10 pays) (MESURE, 0.93)
- Work Trend Index 2026 —mesure→ 26 % voient leur direction alignée durablement sur l'IA ; 19 % des utilisateurs en zone Frontier (MESURE, 0.92)
- article La fin de Valeur × Faisabilité —est_basé_sur→ Work Trend Index 2026 (DOCUMENT, 0.92)
- article La fin de Valeur × Faisabilité —référence→ One New Zealand (ORGANISATION, 0.88)
- Nicolas Suchaud —recommande→ remplacer le backlog de use cases par un portefeuille de décisions métier (AFFIRMATION, 0.95)
- Nicolas Suchaud —recommande→ ne construire que ce qui différencie et garder de la réversibilité sur le reste (AFFIRMATION, 0.93)
- Nicolas Suchaud —recommande→ choisir le mode de travail avec l'agent avant de choisir l'outil (AFFIRMATION, 0.92)
- Nicolas Suchaud —affirme_que→ on ne priorise plus ce qu'on peut faire, on priorise ce qu'on accepte de déléguer (CITATION, 0.95)

---
Canonical: https://www.thekb.eu/en/fiches/suchaud-wenvision-fin-valeur-faisabilite-prioriser-ia-2026-10-22/