Procurement / Sourcing
Procurement buys what the company needs at the best total cost, and manages supplier relationships and risk.
Key processes
- Procure-to-pay
- Supplier selection
- Contract negotiation
- Inventory management
- Supplier risk
KPIs that matter
- Cost savings
- Negotiated savings vs. baseline prices.
- Supplier on-time delivery
- Share of orders delivered on time.
What AI agents can do here
- Spend analysis
- Invoice matching
- Supplier research
- Autonomous tail-spend negotiation
- Supply-risk monitoring
- Agents negotiating with supplier agents
What good looks like
- Three-way match: purchase order, goods receipt and invoice must agree before payment.
- Dual-source critical inputs and track supplier risk.
- Negotiate framework contracts for recurring spend.
Common pitfalls
- Single-source dependencies on critical inputs
- Maverick spend outside contracts
- Accepting unfavorable payment terms under time pressure
Agent tasks and the human control they need
| Task | Maturity | Human control |
|---|---|---|
| Spend analysis and supplier research | Today | Category managers validate recommendations. |
| Negotiating small, standard purchases (tail spend) within limits | Emerging | Spend and term limits; humans approve new suppliers. |
| Agents negotiating with supplier agents | Future (3–5 years) | Humans own strategic suppliers and contract terms. |
Today: in common production use. Emerging: piloted by early adopters. Future: plausible within 3–5 years, not yet dependable.
End-to-end processes this department is part of
Related processes
Procure-to-pay
Purchase request → Supplier selection → Purchase order → Goods receipt → Invoice matching → Payment
Three-way matching (PO, receipt, invoice) is one of the first finance tasks companies automate with AI.
Sales & operations planning (S&OP)
Demand review → Supply review → Reconcile gaps → Executive decision → Publish plan
Without S&OP, Sales sells capacity that doesn't exist, and Operations builds stock nobody ordered.
Plan-to-produce
Demand plan → Material requirements → Capacity planning → Schedule → Produce → Quality check
Predictive maintenance agents can flag failing machines weeks before breakdown.