Analysis of prompts from a Generative AI pilot
The AI team of the municipality of Amsterdam is experimenting with generative AI to advise on its practical use within municipalities.
This document presents the findings of an analysis of a pilot where 150 civil servants used a GPT-based generative AI chatbot for four weeks. This study focused on prompt evaluation, analyzing user behavior, bias, and factuality. Additionally, based on the analysis findings, risk mitigation strategies are proposed.