Researchers built PolicyGuide, a system that compiles organizational policies into workflow graphs and uses a proactive verifier at user-turn boundaries to steer customer-service LLM agents toward compliant behavior, addressing failures where agents either perform forbidden actions or skip required steps. Tested across airline, retail, and telecom domains, the approach improved average compliance rates from 0.42 to 0.62, with especially strong results in structured workflow environments. The system also transferred successfully across multiple underlying LLMs.
