From hype to utility
The rapid emergence of new artificial intelligence tools is reshaping how development organisations operate. For EOF, the question has not been whether to adopt AI, but how to do so in a way that strengthens – not distorts – its core mission: improving outcomes for children and young people.
Our approach has been pragmatic. Rather than pursuing AI as an end in itself, it has been applied to solve operational challenges, improve efficiency, and enhance how programmes are designed, delivered and evaluated. This has been made possible by a distinctive asset: a growing body of high-quality, real-world data generated through outcomes-based programmes across countries, partners and sectors.
Three practical applications
Three applications illustrate this approach.
First, EOF has developed a knowledge agent to unlock organisational data. Large volumes of qualitative information – programme documents, guidance and research – are often underused. By mapping, securing and integrating these datasets into a searchable AI interface, staff can access institutional knowledge more easily, improving consistency and decision-making across teams.
Second, the organisation has used large language models to automate cost classification. Analysing financial data across programmes is typically labour-intensive and inconsistent. AI tools now classify transaction data quickly and with greater accuracy, reducing manual workload and creating a cleaner, more comparable dataset. This improves transparency and enables better value-for-money analysis.
Third, EOF is developing a unified quantitative model that integrates costing, target-setting and performance analysis. This model supports programme design and assessment within a single framework, enabling faster decisions, clearer benchmarks and more systematic learning over time.
Keeping control
Across these applications, a consistent set of principles has emerged. Control, transparency and modularity are central. AI systems must be designed so that organisations retain oversight of data, can adapt tools over time, and avoid dependency on single providers. Just as importantly, human judgement remains at the core: AI is used to augment decision-making, not replace it.
A tool, not a solution
The broader implication is not that AI will transform development on its own. Rather, when applied deliberately, it can strengthen the systems that underpin outcomes-based financing – improving the flow of information, reducing inefficiencies, and supporting better decisions.
In a context of rising demand and constrained resources, this offers a practical way to increase both effectiveness and accountability.