We build an AI platform for the World Bank’s communications and business intelligence team to automate open-ended survey analysis with AI agents.

The World Bank commissioned Research for Purpose to automate the analysis of open-ended survey answers, for 30 country surveys.

Research for Purpose developed the platform architecture including the features needed to set-up the survey question, answers, categories needed to be analysed by AI. Then we tested AI prompts used by the World Bank and improved it significant, using quality control mechanisms with confidence level metrics and AI agents.

As a result, the World Bank can analyse and segment its data with more accuracy than human analysis (over 95% accuracy across AI categorization). The platform can also segment survey answers according to any filters available in the data and provide an AI-generated summary.

This platform not only saves the World Bank’s analysts tens of hours of manual work but also improves the accuracy of their analysis.