AI-Assisted Country Programme Evaluation

UNICEF regularly commissions evaluations of country programmes to inform the design of its future initiatives. rowsquared was selected to evaluate country programmes in Armenia, Azerbaijan, Georgia, and Kazakhstan based on our deep knowledge of UNICEF’s internal data systems and our proposal to use an innovative, standardized approach to data analysis.

Our methodology combined multiple data sources: internal and external databases and documents, key informant interviews with UNICEF staff and partners in each country, and social media listening. This comprehensive approach allowed us to develop a complete picture of programme performance.

At the core of our work were dynamic, AI-generated reports that summarised and visualised data from the multitude of sources in an accessible way. To achieve this, we developed and orchestrated a comprehensive set of pipelines to extract, process and analyse the input data. By applying advanced AI techniques—classification, summarization, and synthesis—we were able to analyze unstructured content at scale, simplify cross-country coordination, and uncover patterns that might otherwise remain hidden.

These AI-assisted reports provided a foundation for our evaluation team, who integrated these insights with findings from key informant interviews and additional desk reviews. To enhance the review process, we developed a custom RAG (Retrieval-Augmented Generation) chatbot that helped our team efficiently navigate and understand complex document collections.

Each evaluation assessed five critical dimensions of UNICEF programming: relevance, effectiveness, efficiency, coherence, and sustainability. We delivered comprehensive reports that were presented to and discussed with the respective Country Offices. These evaluations provided actionable recommendations that directly informed the development of new country programmes across all four countries.