Recommendation Agent
max +6 bonusUpgrade the current rec system to intelligent, explainable supplier recommendations grounded in FM spend history and supplier performance data. It can gather live data from the internet, build supplier advanced profiles, utilise LLM for data generation and validation, utilise advanced ranking criteria and algorithms. Can support complex input formats like any JSON structure.
North Star Metric
100% of relevant suppliers returned for a given search, with a clear explanation of the decision tree.
Bonus Criteria (+6 pts)
Accuracy vs baseline on real FM data
Compare new rec engine against current system on same historical events. Measurable lift shown side-by-side.
Verified by: Side-by-side output during demo
Uses spend history + performance data
Recs informed by PO history, on-time delivery, price variance — not just category tags.
Verified by: Show data inputs during demo
Explains the recommendation
Agent gives a reason — not just a ranked list. Buyer can understand why.
Verified by: Live demo: ask "why this supplier?"
Rejection feedback loop
When buyer rejects a rec, agent learns or logs why. Even basic logging counts.
Verified by: Show the feedback capture