Procurement Policies

Acemoglu, Autor, and Johnson argue that healthcare and education are the two sectors where pro-worker AI has the greatest potential, and where government has the most leverage. Public funding accounts for roughly 43% of U.S. healthcare spending and about 92% of K-12 education expenditure, giving policymakers significant influence over the direction of AI adoption. They propose using procurement, reimbursement policy — such as Medicare and Medicaid payment rules for AI-assisted care delivered by non-physician professionals — and public R&D funding to steer innovation toward tools that complement rather than replace workers. The authors also advocate DARPA-style challenge prizes as an alternative to traditional procure-to-specification contracts, encouraging developers to compete to build AI systems that expand worker capabilities.

On procurement, the Plan directs the General Services Administration (GSA) to create an "AI procurement toolbox" enabling uniform adoption across agencies to the greatest extent practicable, establishes an "Advanced Technology Transfer and Capability Sharing Program" for rapid cross-agency capability transfer, and calls for piloting AI to improve public service delivery through "High Impact Service Providers."

The UK's comprehensive guidance for public sector AI adoption establishes ten principles governing responsible government use, with detailed procurement guidance covering routes to market (frameworks, Dynamic Purchasing Systems), specification requirements, and alignment with data ethics frameworks. It addresses vendor lock-in, intellectual property rights, bias mitigation requirements, and ongoing support obligations. The Playbook's emphasis on the Algorithmic Transparency Recording Standard (ATRS) for public disclosure and its cautionary guidance on "use cases to avoid" (including fully automated decision-making in high-risk areas) offers a useful comparative framework to the current U.S. approach, which prioritizes acceleration over precaution.

This analysis identifies three key trends in public sector AI procurement: the dominance of off-the-shelf licenses over custom builds for efficiency-driven use cases, the shift toward centralized enterprise-wide procurement, and the rise of "shadow AI" entering government through free pilots, grants, and features embedded in existing tools that bypass formal procurement oversight. The report recommends that governments develop comprehensive AI strategies identifying high-ROI use cases (citing Seattle's 2025-2026 AI Plan as a model), build in-house technical capacity to evaluate vendors and manage deployments, and use procurement as a "driver of innovation" rather than passive purchasing.