The Procurement AI Reality Check
Once a month we hand the same procurement task to ChatGPT, Claude, Gemini and Copilot and publish what they caught and what they missed. Every finding comes from a real platform session on a dated run, not a demo.
The task is a realistic case file — a SaaS renewal, a freight contract, a landed-cost quote — the kind of document that lands on a category manager’s desk with a “can you check this before the meeting?” attached. The case data is synthetic; the traps in it are the ones real files carry.
Nothing in it is a benchmark score. It is a field note: here is what happened when a working procurement person asked a general-purpose AI to do a working procurement job. We publish the misses because the misses are where the money is.
Issue 1 — July 2026
The first issue publishes on this page on Monday, 6 July 2026, alongside the first official PAIR-20 run it draws on.
Don’t wait for the monthly issue
The course already shows twenty of these tasks worked end to end, with the platform evidence on the page. Lesson 2 is free to read.