Sélection et achats de fournisseurs IA

Ce qu'elle couvre

Ce programme apprend aux équipes achats et aux responsables de la DSI à structurer l'évaluation des fournisseurs IA, de la première prise de contact jusqu'à la signature du contrat. Les participants apprennent à construire des modèles d'appel d'offres robustes, à définir des critères d'évaluation pertinents, à analyser les clauses de gouvernance des données et de propriété intellectuelle, à réaliser des benchmarks comparatifs et à identifier les signaux d'alerte dans les propositions fournisseurs. Le format associe enseignement structuré et ateliers pratiques sur des documents fournisseurs réels avec des exercices de notation en direct. À l'issue de la formation, les équipes repartent avec des outils réutilisables — modèles d'appel d'offres, grilles de notation et bibliothèques de clauses de données — prêts à être déployés dès le prochain cycle d'achat.

À l'issue, vous saurez

• Draft a complete, AI-specific RFP with weighted evaluation criteria ready to send to vendors
• Apply a structured scoring scorecard to compare at least three competing AI vendor proposals
• Identify and redline problematic data ownership, model training, and IP clauses in vendor contracts
• Design a benchmarking protocol using representative business data to stress-test vendor claims
• Recognise five common red flags in AI vendor demos and proposals and articulate business risk for each

Sujets abordés

• Building AI-specific RFP templates and scoring rubrics
• Defining evaluation criteria: capability, scalability, security, and support
• Data ownership, data processing, and IP clause negotiation
• Benchmarking AI vendor performance on representative tasks
• Identifying red flags in vendor proposals and demo environments
• Vendor risk assessment: financial stability, model drift, and lock-in risk
• Regulatory compliance checks (GDPR, AI Act readiness)
• Onboarding and SLA design for AI-powered services

Modalité

Delivered as a 2–3 day in-person or virtual instructor-led programme, typically split into four half-day sessions to accommodate busy schedules. Each session combines 40% structured instruction with 60% hands-on exercises using anonymised real vendor proposals and contract excerpts. Participants receive a toolkit package including editable RFP templates, a vendor scorecard spreadsheet, a data clause library, and a benchmarking playbook. Remote delivery uses collaborative tools (Miro, shared Google Workspace) to replicate workshop dynamics. On-site delivery recommended when the cohort spans procurement, IT, and legal functions together.

Ce qui fait que ça marche

• Cross-functional cohort (procurement + IT + legal + business owner) completing the programme together to align on criteria before any vendor conversation begins
• Using real, anonymised vendor proposals from past or ongoing evaluations as workshop material to ground learning in lived context
• Establishing a standing AI vendor register and scorecard that is updated after every evaluation cycle to build institutional knowledge
• Engaging legal counsel early to co-own the data and IP clause library rather than treating it as a procurement-only deliverable

Erreurs fréquentes

• Using generic software RFP templates that omit AI-specific criteria such as model explainability, retraining rights, and data residency
• Evaluating vendors solely on demo performance without designing representative benchmark tasks from actual business data
• Signing contracts that grant the vendor rights to use client data for model training without explicit opt-out clauses
• Neglecting lock-in risk by not assessing data portability, API standardisation, and exit clauses before signing

Fournisseurs à considérer

• Wavestonewww.wavestone.com →
• Bearing Pointwww.bearingpoint.com →
• Gartner Learning & Developmentwww.gartner.com/en/training →
• IACCM / World Commerce & Contractingwww.worldcc.com →

Sources

• EU AI Act — Requirements for High-Risk AI System Providers and Users →
• Gartner — How to Evaluate AI Vendors →