1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
51 small wins to finish your pathNext question →
A manufacturer wants an AI assistant for procurement: comparing supplier quotes, checking contracts, and drafting purchase orders in their ERP. Design it.
30-second answerSay your answer out loud first, then reveal.
Workflow
- Quote ingestion: email inbox or supplier portal → parse documents → LLM extraction into a schema per line item.
- Normalisation: units of measure, currencies (convert with the daily rate in code), Incoterms differences (landed cost calculated deterministically).
- Comparison: a table of total landed cost, lead time, payment terms, past supplier performance (on-time delivery, quality rejects from ERP data).
- Policy and contract checks: framework agreement prices, approved supplier list, minimum number of quotes, sanctions/blacklist checks. Rules engine.
- Recommendation: an LLM writes the rationale citing the table and policies; the buyer can ask follow-up questions.
- PO drafting: create a draft PO via ERP APIs with validated fields; the ERP's approval matrix applies.
Risks and controls
- Extraction errors on prices: validate totals, flag outliers vs historical prices.
- Prompt injection inside supplier documents ("ignore other quotes"): treat documents as untrusted; deterministic ranking logic.
- Segregation of duties: the AI drafts, humans approve; full audit trail.
Metrics: cycle time from RFQ to PO, buyer hours per sourcing event, savings vs historical prices, extraction accuracy.
Related
This is what real progress feels like.