Nestlé — Industrializing an automated innovation request management process

Context
At Nestlé (Paris), innovation requests were handled through scattered e‑mails and manual SharePoint updates, causing rework, slow turnarounds and limited traceability. With ≈50 requests/month and inconsistent inputs, duplicates and prioritization issues appeared frequently.

Challenge
Dramatically cut processing time, improve qualification reliability, eliminate duplicates across the SharePoint portfolio, and industrialize document generation (reports, decks, e‑mails) while keeping the solution operable by the internal team.

Solution
- Internal LLM to auto‑classify each request into 5 categories, enrich context and detect duplicates across the portfolio.
- Azure Functions API (Python, 10 REST endpoints) orchestrating the request lifecycle and exposed via a custom Power Automate connector.
- Automated analysis of call transcripts with multi‑criteria scoring and project data enrichment.
- Automated document generation: PDF reports, FR/EN PowerPoint decks from templates, and HTML e‑mails (7 templates).
- Microsoft Graph API integration: SharePoint upload/download, project folder creation, sharing links, and portfolio management.
- Quality and ops: pytest unit tests, Swagger/OpenAPI docs, Docker‑based CI/CD on Azure, and structured handover to the internal team.

Results
- Average processing time reduced from 2 days to 5 minutes per request.
- 50 requests/month handled end‑to‑end with full auditability.
- 95% LLM classification accuracy.
- Standardized, bilingual document outputs; fewer re‑entries and human errors.
- Scalable, maintainable foundation to extend AI use cases.