2026-10-01T15:55:32Z https://dcpapers.dublincore.org/oai

oai:dcpapers.dublincore.org:952675403 2026-09-10 dcmi-2026 openaire

ImpactAI: Embedding Generative AI into Evidence-Informed Development Decision-Making Wang, Linxi Bassetti, Madeline Lorenzo, Abelardo Zimmer, Philipp Garg, Satvik Piriyatamwong, Piriyakorn Martinelli, Giuliano Hussain, Saqib Sajid, Nihaa Gall, Clémence Khan, Rida Aggarwal, Aarushi Mrema, Jackson Sekwao, Grace Prakash, Jay Sanghavi, Dhruti Bentil, Shadrack Sharifkazemi, Mohammadmehdi Pascale, Sophie Polles, Flavia Fraiberger, Samuel Development practitioners increasingly face the challenge of navigating a rapidly expanding evidence base while making timely decisions under operational constraints. This paper introduces ImpactAI, a generative AI–enabled evidence platform developed within the World Bank’s Development Economics unit to support evidence-informed decision-making in development operations. ImpactAI translates natural-language questions from operational staff, researchers, and policymakers into structured re-trieval and synthesis tasks over a curated corpus of impact evaluations, systematic reviews, and project documents, returning source-grounded responses that summarize what is known, where evidence is strongest or weakest, and how findings may apply to specific contexts. We describe the system’s design principles (evidence grounding, traceability, and workflow alignment), its primary use cases (project preparation, comparative effectiveness analysis, theory of change development, and monitoring and evaluation support), and progress made in 2025 toward institutional deployment, including approval as one of the first generative AI tools positioned for launch within the World Bank and the release of the platform for external user registration. We close with lessons learned on responsible deployment, governance, and managing user demand, and outline next steps for scaling access, improving evidence coverage, and strengthening evaluation of tool performance. Dublin Core Metadata Initiative 2026-09-10 info:eu-repo/semantics/conferenceObject Text application/pdf text/html https://doi.org/10.23106/dcmi.952675403 https://dcpapers.dublincore.org/article/952675403 Dublin Core Metadata Initiative Conference Proceedings eng info:eu-repo/semantics/openAccess https://creativecommons.org/licenses/by/4.0/