AI procurement in the public sector: three findings from our survey
A survey of procurement officers across 14 jurisdictions finds appetite for AI tools running well ahead of the risk-assessment capacity needed to buy them safely.
Between October 2025 and February 2026 we surveyed procurement officers in 14 jurisdictions about how their agencies evaluate artificial-intelligence vendors. Three findings stand out.
First, most agencies report pressure to adopt AI tools quickly, often citing efficiency targets set elsewhere in government, with comparatively little matching pressure to document the basis for vendor claims.
Second, fewer than a third of respondents said their agency had a standing technical capacity to test vendor performance claims independently, relying instead on vendor-supplied benchmarks or reference-customer testimonials.
Third, jurisdictions that had adopted a standard pre-procurement risk-assessment template reported materially fewer post-deployment complaints, though the sample is not large enough to establish causation. The full questionnaire and response tables are published alongside this note.