Client
Industry
Duration
Delivered by
The Project
Architected an AI-powered automated procurement pipeline using n8n and OpenAI to rank multi-vendor products by quality, price, and availability.
- 5x faster cart completion — from hours to under 15 minutes per procurement cycle
- 80% reduction in manual procurement effort across the operations team
- AI scoring consistently outperforms manual selection on cost-quality balance (validated by ops team)
- Handles 500+ product comparisons per procurement run
- Operations team retrained to manage the pipeline in under 2 days
What We Solved
McGrocer's procurement team was manually comparing products across dozens of vendors every day — a process that took hours and produced inconsistent quality decisions based on whoever was doing the work. They needed a system that could ingest multi-vendor product catalogs, score items by quality, price, and availability, and automatically build optimized procurement carts without human intervention.
We designed an event-driven pipeline using n8n as the orchestration layer and OpenAI GPT-4 for multi-criteria product ranking. Vendor catalogs are ingested via API connectors, normalized into a canonical schema, then scored by the AI model against configurable procurement policies. AWS SQS handles the async job queue for large catalog refreshes, and Redis caches scoring results to avoid redundant API calls. The final cart is assembled automatically and presented to procurement staff for one-click approval.
How We Built It
Tech Stack Used
Related Projects
