Improving AI Agent for Public Policy Intelligence and Analysis

Challenges

A civic tech company providing legislative tracking and policy intelligence had built an internal AI Agent to help customers interact with large volumes of U.S. bills and regulations. While the system functioned, it struggled with inconsistent answer quality, slow performance, and vague feature planning around personalization and bill tracking.

Solutions

Gradient Ascent delivered an AI Coaching engagement that addressed both product and technical challenges. We:

Helped define clear processes for testing and evaluating output quality
Facilitated collaboration between product and dev teams to close feedback loops
Guided feature planning for advanced capabilities like alerts and personalized recommendations
Conducted code reviews and data pipeline analysis to identify and implement high-impact improvements

Results

By implementing our recommendations, the client quickly saw improvements in answer quality and performance. Their teams gained a stronger grasp of how to plan, evaluate, and evolve AI Agent features while laying the groundwork for continued innovation.