challenge & innovation

At a federal agency, assessing the effectiveness of anti-harassment policies required a rigorous, evidence-based analysis, but the complexity of language variations and nuanced meanings made manual policy evaluation inefficient. To address this, Savan developed an AWS AI-powered document search and classification tool using Python and UiPath Intelligent Automation. Leveraging deep learning and transformer-based machine learning models, the solution accurately processed synonyms, grammar variations, and subtle contextual differences, enabling more precise policy evaluation and strengthening the agency’s ability to make data-driven, evidence-based decisions.

impact

By extracting insights from existing data sources, including the agency’s administrative records and public-facing websites, Savan minimized the burden on external actors while maximizing analytical depth. The AI-powered tool identified references to the agency’s Anti-Harassment Terms in online publications from educational institutions and assessed their statistical representativeness. This approach achieved 89% accuracy with over 90% precision, enabling the agency to make more informed, data-driven policy decisions without disrupting external stakeholders.