Generating and Evaluating Sustainable Procurement Criteria for the Swiss Public Sector using In-Context Prompting with Large Language Models

A team of Swiss researchers has developed a novel AI system that tackles the complex challenge of translating high-level sustainability regulations into concrete procurement criteria for the public sector. Their configurable pipeline uses in-context prompting with large language models (LLMs) to ingest official Swiss government and European Commission sustainability guidelines as structured reference documents. The system then systematically generates verifiable, sector-specific criteria catalogs covering selection criteria, award criteria, and technical specifications—tasks that typically require substantial manual effort and domain expertise across numerous goods and service categories.

The research paper demonstrates the system as a proof-of-concept software tool that integrates interchangeable LLM backends with automated output validation, creating an auditable generation process. Evaluation involved both automated quality checks (including an LLM-based evaluation component) and expert comparison against manually curated gold standards. Results show the pipeline can significantly reduce manual drafting workload while producing criteria consistent with official guidelines. The researchers also document important limitations, failure modes, and design trade-offs observed during deployment, providing crucial insights for integrating generative AI into public sector software workflows where accuracy and auditability are paramount.