implementation-readiness-report-2026-02-27.md 44 KB


stepsCompleted: ["step-01-document-discovery", "step-02-prd-analysis", "step-03-epic-coverage-validation", "step-04-ux-alignment", "step-05-epic-quality-review", "step-06-final-assessment"] documentInventory: prd: prd.md architecture: architecture.md epics: epics.md

ux: ux-design-specification.md

Implementation Readiness Assessment Report

Date: 2026-02-27 Project: agentMarketing


PRD Analysis

Functional Requirements

Workspace & Account Management

  • FR0: Le Transformateur peut s'inscrire avec email + mot de passe, confirmer son email via un lien envoyé par mail, puis se connecter (décision 2026-02-27 : pas d'OAuth)
  • FR1: Le Transformateur peut créer un workspace en renseignant le nom de l'entreprise et son email
  • FR2: Le Transformateur suit un onboarding guidé à la création du workspace (checklist : configurer le contexte entreprise → soumettre une 1ère idée → inviter des contributeurs → attendre les premières soumissions)
  • FR3: Le Transformateur peut configurer les informations de base de son workspace (nom, secteur d'activité, taille de l'entreprise) — ces données alimentent la personnalisation des exemples IA et des suggestions de la bibliothèque
  • FR3b: Le système utilise le secteur et la taille configurés pour contextualiser les appels LLM ; si secteur non configuré, le LLM utilise un contexte générique entreprise PME
  • FR4: Le système applique les limites du tier Free (nombre d'idées, contributeurs, exports)
  • FR5: Le Transformateur peut voir son niveau d'abonnement et les limites associées

Contributor Management

  • FR6: Le Transformateur peut inviter des contributeurs par email
  • FR7: Le contributeur invité reçoit un email avec un lien direct vers le formulaire de soumission
  • FR8: Le contributeur peut accéder au formulaire et soumettre une idée sans créer de compte
  • FR9: Le Transformateur peut voir la liste des contributeurs invités avec leur statut (invité / a soumis / n'a rien soumis)
  • FR10: Le système envoie automatiquement une relance aux contributeurs sans soumission 21 jours après leur invitation
  • FR11: Le Transformateur peut envoyer manuellement une relance à tous les contributeurs du workspace
  • FR12: Le Transformateur peut personnaliser le message de la relance manuelle (objet + corps) avant envoi
  • FR13: Le Transformateur reçoit une confirmation avant envoi d'une relance manuelle indiquant le nombre de destinataires

AI Education & Use Case Library

  • FR14-IA1: À l'ouverture du formulaire, le système affiche 2-3 exemples de cas d'usage IA contextuels générés par le LLM (secteur + taille workspace + département contributeur si connu)
  • FR14-IA2: Les exemples affichés sont rédigés en langage métier sans acronymes ni termes techniques IA (NLP, OCR, ML, LLM, API, etc.)
  • FR14-IA3: Le contributeur peut accéder à la bibliothèque de cas d'usage IA depuis le formulaire ; la bibliothèque est organisée par secteur et par fonction
  • FR14-IA4: Le contributeur peut sélectionner un cas de la bibliothèque pour pré-remplir le formulaire (titre + description partiellement renseignés)
  • FR14-IA5: Le CDO peut accéder à la même bibliothèque depuis son tableau de bord et partager un lien direct vers un cas ou une catégorie
  • FR14-IA6: La bibliothèque contient au minimum 5 cas d'usage IA par secteur principal (distribution, industrie, services, santé, finance) et par fonction principale (RH, finance, supply chain, commercial, opérations) au lancement

Idea Submission

  • FR14: Le contributeur peut soumettre une idée via formulaire web (5 champs : description du problème, département, impact estimé, fréquence, pièces jointes optionnelles)
  • FR15: Le contributeur peut soumettre une idée par email à une adresse inbox dédiée par workspace
  • FR16: Le contributeur reçoit une confirmation visuelle de soumission
  • FR17: Le contributeur peut joindre des fichiers (PDF, DOCX, XLSX, PNG, JPG ; max 10 Mo/fichier ; max 3 pièces jointes)

AI Enrichment & Scoring

  • FR18: Le système enrichit automatiquement chaque idée soumise via LLM (asynchrone, < 10s) : titre normalisé (< 8 mots), domaine métier (RH/Finance/Supply Chain/Commercial/Opérations/IT/Autre), type d'IA pressenti, explication du potentiel IA en 2-3 phrases non-techniques, scores suggérés (impact, faisabilité, urgence sur 1-5)
  • FR18b: L'explication LLM est directement lisible par le contributeur et le CDO — sans jargon, décrivant ce que l'IA ferait concrètement et les données requises
  • FR19: Le Transformateur peut voir les scores LLM et l'explication enrichie dans la fiche détaillée et dans la vue backlog
  • FR20: Le Transformateur peut ajuster manuellement les scores (1-5) — la valeur ajustée remplace la suggestion LLM et est tracée dans l'historique
  • FR21: En cas d'échec LLM, la fiche est créée sans enrichissement en ≤ 30s ; un indicateur "À qualifier manuellement" est affiché ; le CDO peut déclencher manuellement une nouvelle tentative

Backlog & Idea Management

  • FR22: Le Transformateur peut voir toutes les idées en vue liste triable (score, date, statut, domaine)
  • FR23: Le Transformateur peut voir toutes les idées en vue kanban 4 colonnes (Idée / À qualifier / Prioritaire / En cours)
  • FR24: Le Transformateur peut filtrer le backlog par statut, domaine et contributeur
  • FR25: Le Transformateur peut ouvrir la fiche détaillée d'une idée (description, score, statut, historique, pièces jointes)
  • FR26: Le Transformateur peut changer le statut d'une idée (Envoyé / En traitement / Mis à l'ordre du jour / En cours / Non prioritaire) — les idées "Non prioritaire" sont exclues de la vue active mais accessibles via filtre
  • FR27: Le Transformateur peut rechercher des idées par mots-clés dans le backlog
  • FR28: Le Transformateur peut voir un résumé des nouvelles activités depuis sa dernière visite

Qualification & Communication

  • FR29: Le Transformateur peut poser une question de qualification à l'auteur d'une idée depuis la fiche
  • FR30: Le contributeur reçoit une notification email quand une question de qualification lui est posée
  • FR31: Le contributeur peut répondre à la question de qualification (depuis email ou plateforme)
  • FR32: Le Transformateur reçoit une notification email à chaque nouvelle idée soumise dans son workspace

Prioritization

  • FR33: Le Transformateur peut réordonner les idées par drag & drop dans le backlog
  • FR34: Le Transformateur peut voir le classement de priorité résultant de ses ajustements

Export & Sharing

  • FR35: Le Transformateur peut générer un export PDF du pipeline d'idées priorisé
  • FR36: Le Transformateur peut générer un lien partageable en lecture seule vers son pipeline

Compliance & Data

  • FR37: Le système affiche une mention RGPD sur le formulaire de soumission avant validation
  • FR38: Le Transformateur peut supprimer des idées et des données contributeurs sur demande

Total FRs: 47 (FR0, FR1–FR38, including FR3b, FR14-IA1 through FR14-IA6, FR18b)


Non-Functional Requirements

Performance

  • NFR1: Formulaire contributeur chargé en < 2s pour 95% des accès
  • NFR2: Confirmation de soumission affichée en < 1s après validation
  • NFR3: Génération des exemples IA contextuels en < 3s ; fallback sur exemples génériques pré-calculés si dépassement
  • NFR4: Enrichissement LLM post-soumission complété en < 10s pour 90% des soumissions ; fallback (fiche sans enrichissement) au-delà de 30s
  • NFR5: Backlog chargé en < 3s pour un workspace ≤ 200 idées (connexion standard 10 Mbps)
  • NFR6: Ingestion email (soumission → idée créée) complétée en < 15 minutes pour 95% des emails

Availability & Reliability

  • NFR7: Disponibilité cible : 99% sur fenêtre mensuelle (hors maintenance planifiée notifiée 24h à l'avance)
  • NFR8: Taux d'échec des appels LLM ≤ 5% sur 7 jours glissants ; au-delà, alerte opérationnelle déclenchée
  • NFR9: En cas d'indisponibilité du LLM, le formulaire de soumission reste fonctionnel (exemples génériques affichés, enrichissement différé) — aucune interruption du flux contributeur

Cost & LLM Scalability

  • NFR10: Coût moyen par idée soumise (appels LLM) ≤ 0,05 € en conditions normales
  • NFR11: Tier Free limité à 30 idées avec enrichissement LLM par workspace ; au-delà, nouvelles soumissions sans enrichissement automatique
  • NFR12: Workspace tier Free limité à 50 appels LLM/mois (toutes sources : enrichissements + ouvertures formulaire) ; limite appliquée côté serveur

Security & Data

  • NFR13: Données de tous les workspaces hébergées exclusivement dans des datacenters en Union Européenne
  • NFR14: Isolation complète des données par workspace, y compris dans les prompts LLM (aucun contexte cross-workspace)
  • NFR15: Pièces jointes stockées chiffrées au repos (AES-256 minimum)
  • NFR16: Communications client-serveur chiffrées en transit (HTTPS/TLS 1.2+)

LLM Output Quality

  • NFR17: Exemples contextuels affichés avec indicateur "générique" si secteur workspace non configuré
  • NFR18: Scores LLM affichés avec label explicite "Suggestion IA — à ajuster" dans l'interface CDO

Total NFRs: 18 (NFR1–NFR18)


Additional Requirements & Constraints

  • Multi-tenant model: Workspace is the tenant unit; full data isolation between workspaces
  • RBAC: Two roles only in MVP — Transformateur (admin) and Contributeur; no intermediate role, no public read-only access
  • Freemium model: Free tier (1 workspace, 30 ideas, 5 contributors, 1 PDF export/month, 50 LLM calls/month); Pro tier (unlimited)
  • Integrations in MVP: Email ingestion (dedicated inbox per workspace), OpenAI API (LLM scoring), PDF export
  • Greenfield SaaS B2B: No existing codebase; desktop-first (not mobile-first); no formal SLA in MVP
  • RGPD constraints: EU hosting mandatory, right to access/deletion (manual in MVP), no cross-workspace data sharing, file encryption at rest
  • Tier Free cost limits: LLM call limits enforced server-side to keep infra costs marginal

PRD Completeness Assessment

The PRD is well-structured and thorough. Key strengths:

  • Requirements are numbered and categorized
  • NFRs cover performance, reliability, cost, security, and LLM quality
  • Fallback behaviors are explicitly defined for all LLM interactions
  • RBAC model is clearly specified
  • Freemium tiers are detailed with cost rationale
  • User journeys are concrete and map to specific features

Potential areas to watch during epic validation:

  • FR14-IA6 (library content at launch) implies pre-seeded content — implementation complexity may be underestimated
  • FR8 (no-account contributor flow) requires careful session/token management not further specified in the PRD
  • FR21 (manual LLM retry by CDO) is mentioned but the UX flow for this is not detailed

Epic Coverage Validation

Coverage Matrix

FR Number PRD Requirement (short) Epic Coverage Status
FR1 Création workspace Epic 1 — Story 1.3 ✅ Covered
FR2 Onboarding guidé Epic 1 — Story 1.5 ✅ Covered
FR3 Configuration workspace (secteur, taille) Epic 1 — Story 1.4 ✅ Covered
FR3b Contextualisation LLM via secteur/taille Epic 1 — Story 1.4 ✅ Covered
FR4 Limites tier Free Epic 1 — Story 1.6 ✅ Covered
FR5 Affichage niveau abonnement Epic 1 — Story 1.6 ✅ Covered
FR6 Invitation contributeurs par email Epic 2 — Story 2.1 ✅ Covered
FR7 Email lien direct formulaire Epic 2 — Story 2.1 ✅ Covered
FR8 Accès formulaire sans compte Epic 2 — Story 2.2 ✅ Covered
FR9 Liste contributeurs + statut Epic 2 — Story 2.9 ✅ Covered
FR10 Relance automatique J+21 Epic 2 — Story 2.10 ✅ Covered
FR11 Relance manuelle Transformateur Epic 2 — Story 2.11 ✅ Covered
FR12 Personnalisation message relance Epic 2 — Story 2.11 ✅ Covered
FR13 Confirmation avant envoi relance Epic 2 — Story 2.11 ✅ Covered
FR14-IA1 Exemples IA contextuels (LLM) Epic 2 — Story 2.3 ✅ Covered
FR14-IA2 Exemples sans jargon technique Epic 2 — Story 2.3 ✅ Covered
FR14-IA3 Bibliothèque accessible depuis formulaire Epic 2 — Story 2.4 ✅ Covered
FR14-IA4 Adoption cas → pré-remplissage formulaire Epic 2 — Story 2.4 ✅ Covered
FR14-IA5 CDO accès bibliothèque + lien partageable Epic 2 — Story 2.4 ✅ Covered
FR14-IA6 Bibliothèque min 5 cas/secteur + 5 cas/fonction Epic 2 — Story 2.4 ✅ Covered
FR14 Formulaire soumission 5 champs Epic 2 — Story 2.2 ✅ Covered
FR15 Soumission par email Epic 2 — Story 2.5 ✅ Covered
FR16 Confirmation visuelle soumission Epic 2 — Story 2.7 ✅ Covered
FR17 Pièces jointes (types, tailles, max 3) Epic 2 — Story 2.6 ✅ Covered
FR18 Enrichissement LLM post-soumission Epic 3 — Story 3.1 ✅ Covered
FR18b Explication LLM en langage non-technique Epic 3 — Story 3.1 ✅ Covered
FR19 Scores + explication enrichie visibles CDO Epic 3 — Story 3.4 ✅ Covered
FR20 Ajustement manuel des scores (tracé) Epic 3 — Story 3.4 ✅ Covered
FR21 Fallback LLM + indicateur + retry manuel Epic 3 — Story 3.1 ✅ Covered
FR22 Vue liste triable Epic 3 — Story 3.2 ✅ Covered
FR23 Vue kanban 4 colonnes Epic 3 — Story 3.3 ✅ Covered
FR24 Filtrage backlog Epic 3 — Story 3.2 ✅ Covered
FR25 Fiche détaillée idée Epic 3 — Story 3.4 ✅ Covered
FR26 Changement de statut + Non prioritaire Epic 3 — Story 3.5 ✅ Covered
FR27 Recherche mots-clés backlog Epic 3 — Story 3.6 ✅ Covered
FR28 Résumé nouvelles activités Epic 3 — Story 3.7 ✅ Covered
FR29 Question de qualification Epic 4 — Story 4.1 ✅ Covered
FR30 Notification email question de qualification Epic 4 — Story 4.1 ✅ Covered
FR31 Réponse contributeur à qualification Epic 4 — Story 4.2 ✅ Covered
FR32 Notification email Transformateur (nouvelle idée) Epic 3 — Story 3.8 ✅ Covered
FR33 Drag & drop priorisation Epic 4 — Story 4.3 ✅ Covered
FR34 Classement de priorité visible Epic 4 — Story 4.3 ✅ Covered
FR35 Export PDF Epic 4 — Story 4.4 ✅ Covered
FR36 Lien partageable lecture seule Epic 4 — Story 4.5 ✅ Covered
FR37 Mention RGPD sur formulaire Epic 2 — Story 2.8 ✅ Covered
FR38 Suppression données sur demande Epic 4 — Story 4.6 ✅ Covered

NFR Coverage in Stories

NFR Description (short) Referenced in Story Status
NFR1 Formulaire < 2s (95%) Not explicitly in any story AC ⚠️ Implicit only
NFR2 Confirmation < 1s Story 2.7 ✅ Covered
NFR3 Exemples LLM < 3s + fallback Story 2.3 ✅ Covered
NFR4 Enrichissement < 10s / fallback > 30s Story 3.1 ✅ Covered
NFR5 Backlog < 3s (≤ 200 idées) Story 3.2 ✅ Covered
NFR6 Ingestion email < 15 min Story 2.5 ✅ Covered
NFR7 Disponibilité 99%/mois Infrastructure — no story ⚠️ Not in any story
NFR8 Taux échec LLM < 5%, alerte No story ⚠️ Not in any story
NFR9 Formulaire fonctionnel si LLM down Story 2.3 (fallback) ✅ Partially covered
NFR10 Coût LLM ≤ 0,05€/idée No story ⚠️ Not in any story
NFR11 Limite 30 idées LLM (tier Free) Story 1.6 (FR4/FR5) implicit ⚠️ Implicit only
NFR12 50 appels LLM/mois tier Free (serveur) Story 1.6 (FR4/FR5) implicit ⚠️ Implicit only
NFR13 Hébergement EU Architecture constraint — no story ⚠️ Not in any story
NFR14 Isolation workspace (incl. prompts LLM) Story 1.2 (schema multi-tenant) partial ⚠️ Partially covered
NFR15 Pièces jointes chiffrées AES-256 Story 2.6 ✅ Covered
NFR16 HTTPS/TLS 1.2+ Infrastructure — no story ⚠️ Not in any story
NFR17 Indicateur "générique" si secteur non configuré Story 2.3 (fallback) partial ⚠️ Partially covered
NFR18 Label "Suggestion IA — à ajuster" Story 3.4 ✅ Covered

Missing Requirements

Critical Missing FRs

None — all 46 PRD FRs are covered in epics and stories.

NFR Coverage Gaps (Notable but expected for infra NFRs)

The following NFRs have no dedicated story or explicit acceptance criteria. This is a common pattern for infrastructure-level NFRs, but some are significant enough to warrant attention:

  • NFR1 (Formulaire < 2s) — No story AC explicitly enforces this. Recommendation: Add as an acceptance criterion to Story 2.2 and Story 2.3.
  • NFR7 (99% availability) — Pure infrastructure SRE target; acceptable to handle via deployment configuration rather than a story.
  • NFR8 (LLM error rate monitoring < 5%, alerting) — No story covers building this monitoring. Recommendation: Add a story or at minimum a technical task in Epic 3 to set up LLM error rate monitoring.
  • NFR10 (Cost per idea ≤ €0.05) — No implementation check. Acceptable as an operational tracking metric post-launch.
  • NFR11 & NFR12 (Tier Free LLM call limits) — Mentioned implicitly via FR4 in Story 1.6, but the server-side enforcement logic (limiting calls at the service layer) is not explicitly covered in any story AC. Recommendation: Story 3.1 should explicitly address the server-side quota check.
  • NFR13 (EU hosting) — Pure infrastructure/deployment config. No story needed.
  • NFR14 (Workspace isolation in LLM prompts) — Story 1.2 covers schema-level multi-tenancy, but the explicit LLM prompt isolation requirement (ensuring no cross-workspace data in prompts) is not in any story AC. Recommendation: Add explicit AC to Story 3.1 (enrichissement) and Story 2.3 (exemples).
  • NFR16 (HTTPS/TLS) — Deployment config only.
  • NFR17 (Indicateur "générique") — Partially in Story 2.3 but not explicit. Recommendation: Add as explicit AC in Story 2.3.

Coverage Statistics

  • Total PRD FRs: 46
  • FRs covered in epics: 46
  • FR Coverage percentage: 100%
  • Total PRD NFRs: 18
  • NFRs with explicit story coverage: 6 (NFR2, NFR3, NFR4, NFR5, NFR6, NFR15, NFR18)
  • NFRs partially covered: 4 (NFR9, NFR11, NFR12, NFR14, NFR17)
  • NFRs not in any story (infra/operational): 4 (NFR1, NFR7, NFR8, NFR10, NFR13, NFR16)
  • NFR explicit story coverage: ~39% (acceptable for a SaaS MVP — infra NFRs typically handled outside stories)

UX Alignment Assessment

UX Document Status

Found — ux-design-specification.md (45 KB, 27/02/2026), complete 14-step workflow.

The UX spec covers: executive summary, target users, design challenges, design system (Tailwind + shadcn/ui + Livinia Chaleureux palette), responsive strategy, accessibility (WCAG 2.1 AA), component library (6 custom components defined), user journey flows (J1, J2, J2b, J3), UX consistency patterns, emotional design, and competitive benchmarking.


A. UX ↔ PRD Alignment

Well-Aligned Areas:

UX Spec Element PRD Requirement Status
DepartmentPicker + ExamplePill avant le formulaire FR14-IA1 (exemples contextuels selon secteur + département) ✅ Aligned
Label "Suggestion IA — à ajuster" systématique NFR18 ✅ Aligned
Fallback LLM immédiat < 3s / < 10s NFR3, NFR4, NFR9 ✅ Aligned
StatusStepper (Envoyé → En traitement → À l'ordre du jour) FR26 (statuts contributeur) ✅ Aligned
Confirmation visuelle < 1s NFR2, FR16 ✅ Aligned
Empty state rassurant (backlog vide J1-J3) PRD Journey 1 (implicit) ✅ Aligned
Drag & drop (dnd-kit) FR33, FR34 ✅ Aligned
Export en position primaire FR35 ✅ Aligned
Desktop-first, mobile form PRD MVP feature #20 ✅ Aligned
WCAG 2.1 AA Architecture NFR constraint ✅ Aligned

UX Additions Beyond PRD — Gaps to Address:

UX Spec Requirement PRD Coverage Gap Level
Email de confirmation au contributeur après soumission (avec CTA "Soumettre une autre idée") Not in PRD FRs. FR16 only covers visual confirmation (in-app). ⚠️ NOTABLE — UX adds an email confirmation not in PRD nor in any story. Should be added as AC to Story 2.7 or as a new story.
Vue "Cartes" (grid cards) comme 3ème vue backlog PRD has FR22 (liste) + FR23 (kanban) — only 2 views. UX direction 2 adds a card grid view. ⚠️ SCOPE ADDITION — Card grid view is defined in UX design directions but no corresponding FR or epic story covers it. Needs resolution: include in Story 3.2 or explicitly defer to post-MVP.
DepartmentPicker à l'ouverture si département inconnu (1 clic avant exemples) FR14-IA1 says "si le département est connu, les exemples sont filtrés par fonction" — implies it might be unknown but doesn't specify the UX flow to ask for it. ⚠️ UX DETAIL — The department picker interaction is not in any story AC. Should be added to Story 2.3.
Rappel J+30 au contributeur ("pas de signal revenir soumettre" mitigation) Not in PRD FRs. Growth features include "notifications email contributeur enrichies" but J+30 reminder is not defined in MVP scope. ℹ️ LOW — Pre-mortem mitigation in UX; not confirmed as MVP scope. Mark as out-of-scope until PRD is updated.
Tooltip sur chaque champ enrichi + indicateur de confiance (pre-mortem mitigation) Partially in PRD: NFR18 (label "Suggestion IA") + FR20 (ajustement possible). Tooltips not explicitly specified. ℹ️ LOW — Implementation detail; acceptable as developer discretion within the spirit of NFR18.

B. UX ↔ Architecture Alignment

Well-Aligned Areas:

UX Requirement Architecture Support Status
Tailwind + shadcn/ui + Livinia Chaleureux Explicitly specified in architecture decision ✅ Aligned
dnd-kit for drag & drop (kanban + backlog) Explicitly listed in frontend architecture decisions ✅ Aligned
DepartmentPicker, ExamplePill, IdeaCard, AIBadge, ScoreSlider, KanbanBoard, EmptyState, StatusStepper Architecture's components/form/, components/backlog/, components/idea/ cover all UX components (some naming differs slightly: ExamplesPanel.tsx vs. ExamplePill) ✅ Aligned (minor naming delta — not a blocker)
2 interfaces asymétric (CDO dense / contributeur épuré) Architecture separation: (dashboard)/ for CDO, (public)/submit/ for contributor ✅ Aligned
LLM fallback (skeleton → generics) Architecture's services/llm.ts with fallback, skeleton pattern documented ✅ Aligned
WCAG 2.1 AA Architecture lists Radix UI under shadcn/ui; ARIA documented in UX spec; explicitly referenced in architecture ✅ Aligned
Route for share link Architecture has app/share/[shareToken]/page.tsx ✅ Aligned
Breakpoints (sm:640, md:768, lg:1024, xl:1280) Architecture references same breakpoints from UX spec ✅ Aligned

Architecture Gaps vs UX:

UX Requirement Architecture Gap Gap Level
Card grid view (direction 2) Not present in architecture's component list or routing structure ⚠️ MINOR — If card view is confirmed in-scope, need to add components/backlog/CardGrid.tsx and update Story 3.2
Email confirmation to contributor post-submission services/email.ts exists but no explicit route/procedure for this email ⚠️ MINOR — Needs explicit addition to email.ts + Story 2.7
LLM error rate monitoring (NFR8) Architecture notes it as a "nice-to-have" — no monitoring service or alerting defined ⚠️ MINOR — Acceptable for MVP launch, but should be planned for early post-launch

Warnings

  1. Email confirmation after submission is a missing story: The UX spec explicitly includes an email to the contributor after submission (with CTA to submit another idea). This is not an FR in the PRD and not in any story AC. Without it, the "boucle de retour" for the contributor is broken. Recommendation: Add AC to Story 2.7 or create Story 2.7b.

  2. Card grid view (Vue Cartes) has no story: The UX spec introduces a third backlog view (card grid). The PRD only defines FR22 (list) + FR23 (kanban). This is a scope addition with no story. Recommendation: Either defer explicitly to post-MVP or add story in Epic 3.

  3. DepartmentPicker flow not explicit in Story 2.3: The UX spec's J2 flow shows that if the department is unknown, a department picker is shown BEFORE examples are generated. This interaction is critical for FR14-IA1 accuracy but not in any story AC. Recommendation: Add explicit AC to Story 2.3.

  4. Architecture's ExamplesPanel.tsx vs. UX's ExamplePill + DepartmentPicker split: The architecture bundles both into one file, while UX spec defines two distinct components. Minor naming inconsistency — no blocker, but developer should follow UX spec decomposition for clarity.

Overall UX Alignment: GOOD — The UX spec is detailed, well-aligned with PRD and Architecture. Three notable additions (email confirmation, card grid view, department picker flow) need explicit story coverage or deferral decisions before implementation starts.


Epic Quality Review

Review Scope

4 epics, 23 stories reviewed against create-epics-and-stories best practices.


Epic 1: Foundation & Workspace — Quality Assessment

User Value Focus: ✅ PASS (with caveat)

  • Epic goal is user-centric: "Le Transformateur peut créer son workspace, configurer secteur/taille et disposer d'une base opérationnelle."
  • Stories 1.1 (scaffold T3) and 1.2 (Prisma schema) are technical/developer stories. However, this is justified and expected for a greenfield project — architecture explicitly mandates Story 1.1 as the starter template story. ✅ Acceptable by convention.

Epic Independence: ✅ PASS — Epic 1 stands alone completely.

Story Dependencies within Epic 1:

Story Dependencies Valid?
1.1 (scaffold) None ✅ Independent
1.2 (schema) 1.1 (T3 project with Prisma) ✅ Valid backward
1.3 (FR1 create workspace) 1.1 + 1.2 ✅ Valid backward
1.4 (FR3 config secteur/taille) 1.3 (workspace exists) ✅ Valid backward
1.5 (FR2 onboarding) 1.3 + 1.4 ✅ Valid backward
1.6 (FR4/FR5 tiers) 1.3 ✅ Valid backward

🟡 Minor Concern — Story 1.5 forward reference: Story 1.5's onboarding checklist includes step "inviter des contributeurs" (Epic 2 feature) and "soumettre une 1ère idée" (also needs Epic 2 form). The checklist UI can be built, and steps will show as "pending" since the underlying features don't exist yet. This is acceptable if the ACs are interpreted as: the checklist renders correctly and marks steps complete when the actions are eventually performed. However, the AC "les étapes sont skipables sauf la configuration de base" implies interaction — which needs the steps to be functional. Recommendation: Add AC clarification that checklist items show "À venir" state until the underlying feature of a later epic is deployed.

Acceptance Criteria Quality — Epic 1:

  • All ACs use Given/When/Then format ✅
  • ACs are specific and testable ✅
  • Error conditions: Story 1.3 does not define what happens if workspace creation fails (e.g., duplicate slug) — minor gap 🟡

Epic 2: Invitation & soumission d'idées — Quality Assessment

User Value Focus: ✅ PASS — Clearly user-centric, delivers core contributor and Transformateur workflows.

Epic Independence: ✅ PASS — Functions using only Epic 1 output.

Story Dependencies within Epic 2:

Story Dependencies Valid?
2.1 (invitation FR6/FR7) Epic 1 workspace ✅
2.2 (form FR8/FR14) Epic 1 workspace ✅
2.3 (examples FR14-IA1/IA2) 2.2 (form exists), Epic 1 LLM config ✅
2.4 (library FR14-IA3 to IA6) Epic 1 workspace ✅
2.5 (email ingestion FR15) Epic 1 workspace, email webhook ✅
2.6 (attachments FR17) 2.2 (form) ✅ Valid backward
2.7 (confirmation FR16) 2.2 (form) ✅ Valid backward
2.8 (RGPD FR37) 2.2 (form) ✅ Valid backward
2.9 (contributor list FR9) 2.1 (invitations) ✅ Valid backward
2.10 (auto relance FR10) 2.1 (invitations) ✅ Valid backward
2.11 (manual relance FR11-13) 2.1 (invitations) ✅ Valid backward

🟠 Major Issue — Story 2.2 forward dependency on Story 2.6: Story 2.2 acceptance criteria explicitly includes "pièces jointes optionnelles" as one of the 5 form fields: "When je remplis les champs (description problème, département, impact estimé, fréquence, pièces jointes optionnelles)." But the actual attachment implementation (upload, file type validation, storage path) is in Story 2.6.

  • Impact: Story 2.2 cannot be completed as specified without Story 2.6's work.
  • Recommendation: Either (a) remove "pièces jointes" from Story 2.2's AC and specify the form has 4 fields initially, then Story 2.6 adds the 5th; or (b) merge Stories 2.2 and 2.6; or (c) reorder so Story 2.6 comes before 2.2.

🟠 Major Issue — Story 2.4 oversized (potential): Story 2.4 bundles 4 FRs (FR14-IA3, FR14-IA4, FR14-IA5, FR14-IA6) including:

  • Building the library UI (browse by sector/function)
  • Implementing library → form pre-fill
  • CDO access + link sharing
  • Creating and seeding ≥5 AI use cases per sector (5 sectors × 5 functions = 25+ use cases) as seed data

The content creation requirement (FR14-IA6) is significant and distinct from the technical implementation. Bundling seed data creation with the technical library feature in one story risks a very large story that could block an entire sprint.

  • Recommendation: Split Story 2.4 into:
    • Story 2.4a: Library UI (browse, filter by sector/function, CDO access, share link) — FR14-IA3, FR14-IA4, FR14-IA5
    • Story 2.4b: Library content seed (create ≥5 use cases per sector + function) — FR14-IA6

🟡 Minor Concern — Story 2.10 settings reference: Story 2.10 AC: "la relance automatique peut être désactivée par le Transformateur dans les paramètres du workspace." No story explicitly creates a workspace settings page with this toggle. The settings page is implied in the architecture (workspace/[workspaceId]/settings/page.tsx) but no story covers it. If Story 2.10 requires this settings toggle to be built, it creates an implicit dependency on an unspecified story. Recommendation: Add a brief story or sub-task within Epic 1 Story 1.4 or Story 1.6 to create the workspace settings page scaffold, or add the relance toggle as an AC to Story 1.6.

Acceptance Criteria Quality — Epic 2:

  • Given/When/Then format: ✅ consistent
  • Story 2.3 has good fallback coverage ✅
  • Story 2.4 AC "pas nécessairement toutes les combinaisons" (for library content) is appropriately nuanced ✅
  • Story 2.11 AC is complete (includes confirmation + log) ✅
  • 🟡 Story 2.5 (email ingestion) doesn't specify what happens if the email cannot be parsed (malformed email body) — missing error condition

Epic 3: Backlog & enrichissement IA — Quality Assessment

User Value Focus: ✅ PASS — Core CDO workflow, clearly user-centric.

Epic Independence: ✅ PASS — Functions using Epic 1 + 2 output (needs ideas submitted).

Story Dependencies within Epic 3:

Story Dependencies Valid?
3.1 (enrichissement FR18/18b/21) Ideas submitted (Epic 2), LLM service ✅
3.2 (liste FR22/24) Ideas exist (Epic 2) ✅
3.3 (kanban FR23) 3.2 or ideas exist ✅
3.4 (fiche FR25/19/20) Ideas exist, enrichment (3.1) ✅ Valid backward
3.5 (statuts FR26) Ideas exist ✅
3.6 (recherche FR27) Ideas exist ✅
3.7 (résumé activités FR28) Ideas exist + ⚠️ ⚠️ See below
3.8 (notification FR32) Ideas exist + email service ✅

🟠 Major Issue — Story 3.7 forward dependency on Epic 4: Story 3.7 AC: "je vois un résumé des nouvelles idées et des réponses reçues aux questions de qualification." The "réponses reçues" part requires the qualification feature from Epic 4 (Stories 4.1/4.2) to exist.

  • Impact: Story 3.7 as written cannot be fully completed without Epic 4 features.
  • Recommendation: Split Story 3.7 into two phases:
    • Story 3.7 (Epic 3): Shows résumé of new ideas only. AC removes "réponses reçues."
    • Story 4.x (Epic 4): Enhances résumé to also show qualification responses. AC adds "réponses reçues aux questions de qualification."

Acceptance Criteria Quality — Epic 3:

  • Story 3.1 ACs are detailed and cover the fallback ✅
  • Story 3.3 provides the kanban column mapping clarification ✅ (excellent AC)
  • Story 3.4 references label NFR18 explicitly in the AC ✅
  • Story 3.7 defines "dernière visite" in the AC ✅ (good precision)
  • 🟡 Story 3.2 doesn't specify behavior when filter returns 0 results (empty filter state)
  • 🟡 Story 3.1 doesn't explicitly cover the LLM prompt isolation requirement (NFR14 — no cross-workspace data in prompts). Recommendation: Add AC "And the LLM enrichment prompt contains only data from the current workspace."

Epic 4: Priorisation, qualification & export — Quality Assessment

User Value Focus: ✅ PASS — User-centric, covers critical CDO and contributor workflows.

Epic Independence: ✅ PASS — Functions using Epic 1 + 2 + 3 output.

🟡 Minor Concern — Story 4.5 scope expansion: Story 4.5 AC adds: "le lien peut avoir une durée de validité configurable: 7 jours, 30 jours (défaut), ou illimitée." FR36 only says "lien partageable en lecture seule" — no mention of configurable expiry. This is a scope addition beyond the PRD.

  • Impact: Not a blocker, but adds implementation scope not validated in PRD.
  • Recommendation: Either confirm this is in-scope (and update FR36 in PRD) or simplify to a fixed duration (e.g., always 30 days) for MVP.

🟡 Minor Concern — Story 4.6 implicit dependency on Story 4.5: Story 4.6 AC: "Le lien de partage est révoqué si le contributeur effacé était le seul contributeur visible sur ce lien." This assumes a share token exists (Story 4.5). Within the same epic, this is an acceptable backward dependency — but should be implemented after Story 4.5.

Acceptance Criteria Quality — Epic 4:

  • Story 4.2 clarifies "MVP : réponse uniquement via lien, pas par email direct" ✅ (excellent precision)
  • Story 4.3 (drag & drop) covers both list and kanban drag behavior ✅
  • Story 4.4 references FR4 limits ✅
  • 🟡 Story 4.1 doesn't specify maximum number of qualification questions per idea (1? multiple?). PRD allows multiple questions (journey describes back-and-forth). The AC doesn't define the limit.

Greenfield Project Checks

Check Status
Story 1.1 = starter template setup ✅ Present and correct
Development environment config ✅ Story 1.1 covers .env.example
CI/CD pipeline setup ⚠️ Architecture has .github/workflows/ci.yml but no story covers setting up CI/CD — minor gap for greenfield
Schema creation per story ✅ Implicit (acceptable)

Best Practices Compliance Summary

Epic Delivers User Value Independent Stories Sized OK No Forward Deps Clear ACs
Epic 1 ✅ (1.1/1.2 technical but justified) ✅ ✅ ⚠️ 1.5 minor ✅
Epic 2 ✅ ✅ ⚠️ 2.4 oversized 🟠 2.2 → 2.6 ✅
Epic 3 ✅ ✅ ✅ 🟠 3.7 → Epic 4 ✅
Epic 4 ✅ ✅ ✅ ✅ (internal only) ✅

Quality Findings by Severity

🔴 Critical Violations

None identified. No purely technical epics with zero user value, no epic-sized stories, no circular dependencies.

🟠 Major Issues (Must Fix Before Implementation)

  1. Story 2.2 ↔ Story 2.6 forward dependency — Story 2.2 ACs reference "pièces jointes optionnelles" but the implementation is in Story 2.6. Fix: Remove attachments from Story 2.2 ACs, make it an explicit prerequisite that Story 2.6 must be completed or merged with 2.2 before Sprint closure.

  2. Story 2.4 oversized — Bundles 4 FRs including significant content creation (≥25 use cases). Fix: Split into Story 2.4a (library technical feature) and Story 2.4b (library content seed).

  3. Story 3.7 forward dependency on Epic 4 — "réponses reçues aux questions de qualification" requires Epic 4 features. Fix: Remove that clause from Story 3.7 AC and add an enhancing story in Epic 4.

🟡 Minor Concerns (Improve Before Implementation)

  1. Story 1.5 — onboarding checklist references Epic 2 capabilities — Add AC: checklist items for future epics render in "À venir" state until the underlying feature is deployed.
  2. Story 2.10 — settings toggle not in any story — Add workspace settings scaffold to Story 1.6 or create a Story 1.7.
  3. Story 2.5 — missing error AC — Add AC for malformed or unroutable email body.
  4. Story 3.1 — LLM prompt isolation not explicit — Add NFR14 AC: enrichment prompt must contain only current workspace data.
  5. Story 3.2 — missing empty filter state — Add AC for zero-result filter behavior.
  6. Story 4.5 — scope expansion (configurable expiry) — Confirm or remove. FR36 doesn't specify expiry.
  7. Story 4.1 — qualification question limit — Define maximum questions per idea (e.g., "any number, one at a time") in ACs.
  8. No CI/CD story — Add Story 1.0 or Developer Task for CI pipeline setup (Greenfield best practice).

Overall Epic Quality: GOOD with addressable issues. No blockers that would prevent starting implementation, but 3 major issues should be resolved before the team begins Epic 2 and Epic 3 stories.


Summary and Recommendations

Overall Readiness Status

✅ READY — with 3 required fixes before Epic 2/3

The project is well-prepared for implementation. The PRD is thorough, the architecture is solid, the UX spec is detailed, and the epics cover 100% of the functional requirements. No critical structural failures were found. However, three major issues in the epic stories must be resolved before implementation of Epics 2 and 3 begins, and several notable UX-to-story gaps need explicit decisions.


Findings Summary

Category Total Issues 🔴 Critical 🟠 Major 🟡 Minor
FR Coverage (vs PRD) 0 0 0 0
NFR Coverage (vs PRD) 7 0 0 7
UX Alignment 4 0 2 2
Epic Story Quality 14 0 3 11
Total 25 0 5 20

Critical Issues Requiring Immediate Action (Before Starting Epic 2+)

Issue 1 — 🟠 Story 2.2 ↔ Story 2.6 forward dependency Story 2.2 ACs include "pièces jointes optionnelles" as a form field, but file upload is implemented in Story 2.6. Story 2.2 cannot be closed as complete without Story 2.6. Fix: Reorder stories (2.6 before 2.2 or merged), or remove attachment field from Story 2.2 ACs and explicitly state it is added via Story 2.6.

Issue 2 — 🟠 Story 2.4 oversized (4 FRs + significant content creation) Story 2.4 bundles the library technical feature (FR14-IA3/IA4/IA5) with seeding ≥5 use cases per sector AND per function (FR14-IA6 — minimum 25+ content items). This is too large for a single sprint story. Fix: Split into Story 2.4a (library UI + pre-fill + CDO sharing) and Story 2.4b (content seed ≥5 cases/sector + ≥5 cases/function).

Issue 3 — 🟠 Story 3.7 forward dependency on Epic 4 Story 3.7 includes "réponses reçues aux questions de qualification" in its AC, which depends on the qualification feature in Epic 4 (Stories 4.1/4.2). Fix: Remove "réponses reçues" from Story 3.7 AC. Add a Story 4.x in Epic 4 that enhances the activity summary to include qualification responses.

Issue 4 — 🟠 UX email confirmation to contributor not in any story The UX spec's J2 flow includes an email to the contributor after submission (with CTA "Soumettre une autre idée"). This is not an FR in the PRD and not in any story. Without it, the contributor's feedback loop is incomplete. Fix: Add AC to Story 2.7 (or create Story 2.7b): "And the contributor receives a confirmation email with a CTA to submit another idea."

Issue 5 — 🟠 UX card grid view (Vue Cartes) has no story The UX spec defines a third backlog view (card grid / direction 2) alongside list and kanban. The PRD only defines FR22 (list) and FR23 (kanban). No story covers this view. Fix: Explicitly decide: (a) add to Story 3.2 as a third toggle option, OR (b) defer to post-MVP with a note in the epics document.


Recommended Next Steps

  1. Fix the 3 story dependency issues in epics.md (Issues 1, 2, 3 above) — these are structural: update Story 2.2 ACs, split Story 2.4, and split Story 3.7. These can be done quickly in the epics document before the first sprint.

  2. Decide on the email confirmation to contributor and card grid view (Issues 4, 5) — these require a PRD/scope decision. If confirmed in-scope, update the PRD FRs, update the relevant stories, and confirm architecture support. If out-of-scope for MVP, add a note in epics.md deferring to post-MVP.

  3. Add explicit CI/CD story or developer task — Story 1.0 "Setup CI pipeline (GitHub Actions)" is missing for the greenfield project. Add as a developer task alongside Story 1.1.

  4. Reinforce NFR coverage in story ACs (5 NFRs with no story coverage) — prioritize:

    • Story 2.2/2.3: Add NFR1 AC (formulaire < 2s)
    • Story 3.1: Add NFR14 AC (LLM prompt isolation — no cross-workspace data)
    • Story 3.1: Add NFR11/NFR12 AC (server-side LLM quota check for tier Free)
  5. Clarify minor AC gaps — Update Story 2.5 (malformed email error case), Story 3.2 (empty filter state), Story 4.1 (qualification question limit), Story 4.5 (confirm or remove configurable expiry).


Strengths to Preserve

  • PRD quality is excellent — 46 FRs numbered, all NFRs defined, user journeys concrete, fallback behaviors explicit. Do not let implementation drift from these definitions.
  • Architecture is clear and complete — T3 stack with clear boundaries, multi-tenant patterns, naming conventions, and anti-patterns documented. Teams must follow createWorkspaceProcedure and the LLM isolation rule without exception.
  • UX spec is differentiated — The design principles (no blank page, asymmetric CDO/contributor UX, AIBadge everywhere) are not just polish — they are the core product differentiation. The AIBadge, DepartmentPicker, ScoreSlider, and EmptyState components are critical path items.
  • Epic 1 foundation is solid — The T3 scaffold, Prisma multi-tenant schema, and NextAuth setup provide a strong starting point.

Final Note

Assessment date: 2026-02-27 Assessor: AI Product Manager / Scrum Master (BMAD Check Implementation Readiness workflow) Project: agentMarketing — AImpact Hub

This assessment identified 25 issues across 4 categories: 0 critical, 5 major, 20 minor. The 5 major issues are all resolvable in 1–2 hours of epics document updates before the first implementation sprint begins. The project has strong documentation quality across all four artifacts (PRD, Architecture, UX, Epics) and is fundamentally ready for a greenfield implementation start.

Proceed to implementation after resolving the 5 major issues. Epic 1 can begin immediately as-is.


Post-Assessment Fixes Applied (2026-02-27)

The following fixes were applied to epics.md following the assessment:

  1. Story 2.2 — Removed "pièces jointes" from ACs; formulaire now 4 champs de base. Story 2.6 adds the 5th champ. Forward dependency resolved.
  2. Story 2.4 — Split into 2.4a (library UI, pre-fill, CDO sharing) and 2.4b (content seed ≥5/sector + ≥5/function).
  3. Story 3.7 — Removed "réponses reçues aux questions de qualification" from ACs. Scope limited to "nouvelles idées" only.
  4. Story 4.7 — Added new story in Epic 4 to enhance activity summary with qualification responses (FR28 complement).
  5. Story 2.7 — Added AC for contributor confirmation email with CTA "Soumettre une autre idée".
  6. Scope Deferred — Added section "Scope Deferred to Post-MVP" documenting the card grid view (Vue Cartes) as explicitly deferred.

Status: All 5 major issues resolved. Implementation can proceed.