Methodology

Most commentary on AI and procurement is speculative. This benchmark takes a different approach: instead of asking "will AI replace procurement?", we asked "what exactly does a procurement professional do, and can AI do each part of it?" The result is a ground-up, task-level assessment of AI readiness across Sourcing and Contracting roles.

We started by breaking each procurement role into its constituent tasks — the discrete, repeatable units of work that make up the role. For each task, we went further: mapping the trigger that initiates it, the inputs required, the step-by-step actions performed, the expected output, and the conditions that define success.

We did not stop at the happy path. For every task, we identified the exceptions — the things that go wrong, why they go wrong, and how a skilled practitioner resolves them. This is where most AI assessments fall short. Exception handling is where human judgment, relationship knowledge, and tacit expertise show up most clearly.

For each task, we assessed two dimensions:

AI Readiness — whether AI can perform the task autonomously, with human oversight, or not at all. We applied this at the task level, not the role level, because AI readiness varies significantly within a single role.

Human Effort — how much practitioner time and expertise the task typically demands. Our scoring reflects this: the benchmark is weighted by effort, so the role-level score reflects real-world time displacement.

Where AI falls short, we recorded why — not just that it falls short. We categorized AI limitations into six ceiling types: Judgment, Relationship, Tacit Knowledge, Context, Legal Accountability, and Creativity.

This MVP covers 5 Sourcing and Contracting roles, comprising 58 tasks and several hundred exception scenarios. Coverage will expand to all 35 procurement roles subject to demand from the practitioner community.