Scaling AI in Public Procurement to Enhance Competition and Performance

Public procurement represents a large share of government spending but often suffers from limited competition and information frictions that disadvantage small and medium enterprises (SMEs). This project evaluates whether artificial intelligence, specifically large language models (LLMs), can improve competition and efficiency in public procurement when AI tools are adopted at scale within public administration and market processes.

The study focuses on Chile’s centralized digital procurement platform, Mercado Público. Building on prior work to standardize procurement data, the project evaluates two AI tools: (1) an assistant for public buyers to help formulate clearer and more competitive requests for proposals, and (2) an assistant for suppliers, especially SMEs, to help interpret procurement requirements, analyze historical prices, and prepare competitive bids. The objective is to increase bidder participation and reduce procurement costs.

The intervention will be evaluated through an auction-level randomized controlled trial across multiple procurement markets. Outcomes include bidder participation, bid prices, bid dispersion, winning margins, SME participation and success, and quality-adjusted procurement costs. The study will generate evidence on whether AI can enhance competition, reduce public spending, and lower barriers to participation for SMEs. The project is implemented in partnership with Chile's public procurement office (ChileCompra) and budget office (DIPRES).