LogoLucyWorks
  • Lessons
  • Answers
  • Pricing
  • About
  • Ask Lucy
  • Book a call
LW·A34·Can AI do this?← All answers

How do you use AI in procurement?

Treat it as a drafting tool, not an oracle: give it a clear brief, let it organize the work, then check every number and clause yourself before it shapes a decision.

The method matters more than the tool. Using AI in procurement well comes down to three moves: tell it exactly what you need and what to watch for, let it draft and organize the work into a table you can check row by row, then verify the few numbers that would change your decision. In the July 2026 PAIR-20 run — six test procurement tasks run on ChatGPT, Claude, Gemini, and Copilot — a clear brief turned vague replies into ranked tables with stated assumptions. It did not make the answers automatically correct.

Most people start using AI at work the same way: one browser tab, one question, no formal program. So the practical question is not whether to use it, but how to use it on a specific file — a quote comparison, a contract review, a list of supplier risks — without letting an unchecked answer shape a real decision.

Start with a clear brief, not a one-line prompt. A brief tells the AI what role to play, what columns to fill in, what mistakes to watch for, what format to use, and what assumptions to show instead of hide. For example: "You are a procurement analyst. Compare these three quotes. Normalize all prices to EUR per unit. Flag any supplier that is missing a delivery date." In the July run, that kind of structure turned vague replies into ranked tables with stated assumptions — the difference between an answer you can inspect and one you have to take on trust.

Use it to draft and organize, not to decide. AI is good at turning a long file into a workable issue list, a side-by-side comparison, or a first-pass list of risks. That first pass saves time and catches the obvious problems. But it is not the one who signs off — you are.

Verify the facts that move the decision. In the July run, tasks that produced tidy tables still carried a missed unit conversion, landed-cost totals (the all-in price including shipping and duty) that did not match the case answer, and a field filled with a guess. Those were single dated runs, not a permanent verdict. A miss on one run might not repeat on the next. That is exactly why the human check belongs after every draft, on units, totals, dates, and blank fields.

This is a method, not a shortcut. The same order — brief, draft, verify — works whichever platform your team uses. In the July run, the platforms differed in which mistake they missed, not in whether they needed checking.

From here: what specific procurement jobs AI handles well, how the platforms compare, and what goes into a structured brief.

Where this comes from

  • PAIR-20 July 2026 run of record — single dated runs, screenshots on file

Last checked Sat Jul 11 2026 00:00:00 GMT+0000 (Coordinated Universal Time). Evidence comes from dated, single-run platform sessions with screenshots on file — read each finding as “this happened,” not “this always happens.”

Work this yourself — from the course

Build the AI-Adoption ROI Case That Survives the CFOBuilds a defensible AI-adoption ROI case that accounts for the hidden review-tax cost of checking AI output, structured to survive CFO scrutiny rather than overstating savings.

Related questions

  • What is a structured brief (vs. a prompt)?
  • What are the real use cases for AI in procurement?
  • Which AI platform is best for procurement work?

See what the platforms caught — and missed

Twenty procurement tasks, four AI platforms, real dated runs. Lesson 2 is free to read, no account needed.

Read the free lessonTraining for your team
LogoLucyWorks

AI training that ends with a working agent.

Product
  • The 20 Lessons
  • Answers
  • Ask Lucy
  • PAIR-20 benchmark
  • Reality Check
  • Pricing
  • Book a call
Company
  • About
Legal
  • Cookie Policy
  • Privacy Policy
  • Terms of Service
© 2026 LucyWorks. All Rights Reserved.