Should-cost analysis is the oldest trick in the strategic sourcing playbook: model what a part should cost from the bottom up — raw materials, labour, overhead, margin — and use that number to negotiate suppliers. It works. It is also slow, expensive, and usually stale the moment it is finished. AI changes all three of those things.
The old way
A category engineer opens a spreadsheet. She pulls a BOM, looks up commodity prices from last quarter's index report, adds assumed labour rates from a regional benchmark, tacks on freight and overhead, and lands on a target price. Two weeks later she has a number. Three weeks after that, PET has moved 8% and the number is wrong.
The AI way
An AI procurement copilot does the same decomposition in seconds and keeps it live. Every commodity input is wired to a market feed. Every labour assumption is regional and dated. Every overhead percentage is sourced. When the underlying markets move, the should-cost number moves with them — and the agent pings you when the gap between should-cost and your supplier's invoice crosses a threshold you set.
Why it changes negotiation
- Every ask is cited. "Resin fell 2.1% over 30 days on CME; we expect a $0.11 per unit reduction." Suppliers respect math they can audit.
- Defence is automatic. When a supplier sends a price-increase letter claiming raw-material pressure, the agent has already checked whether the cited commodity actually moved.
- Coverage scales. A team of three can hold sharp positions on thousands of SKUs, not dozens.
What good looks like
A good AI should-cost agent shows its work: cited indices, date ranges, weighting assumptions, and the exact gap to supplier price. It hands those briefs to your buyers in Slack or Microsoft Teams, where the work actually happens. And it keeps learning from how your suppliers respond, so next quarter's model is sharper than this one's.