How can AI help us improve aid effectiveness?
A blog by Seb Mhatre, FCDO pioneer for our pilot ‘Using LLMs as a tool for international development professionals’
Image credit: iStock
AI is changing international development. The biggest effects will be in the digital transformation of developing countries and how that affects their pathways to growth, ability to deliver services for their citizens and the new risks they will face. However, another important way that AI will change international development is the digital transformation of how development itself is delivered. In this blog, I want to talk about the new AI tool we have developed, DevExplorer, as an example of how AI might improve aid effectiveness.
In 2011 in Busan, 161 countries signed up to four principles of aid effectiveness[1]. The four principles are:
Ownership of development priorities by developing countries.
Focus on results.
Inclusive development partnerships.
Transparency and accountability to each other.
These principles are as relevant as ever in 2025. However, the inability by key development actors to access the right knowledge, inexpensively, when they need it is a continuing barrier to aid effectiveness. This is true across the four principles.
Ownership of development priorities by developing countries: A basic prerequisite for ownership of development priorities by developing countries is for developing countries to know what development activities are taking place in their countries.
Focus on results: A focus on results requires development actors to be able to track what results are being delivered and to successfully learn lessons about what works.
Inclusive development partnerships: Creating inclusive development partnerships requires coordination which in turn requires an understanding of who is doing what, where, and with whom.
Transparency and accountability to each other: Accountability requires that all stakeholders can know what development activities are taking place.
The first step in solving these challenges requires a system for collecting data about development activities that ensures that the data is relevant, high quality, accessible and regularly updated. Thankfully, a system with the goal of collecting such data is already in place. The International Aid Transparency Initiative (IATI) “brings together governments, multilateral institutions, private sector and civil society organisations and others”[2] with the mission[3] to:
ensure transparency of data on development resources and results;
ensure the quality of IATI data is continually improved and responds to the needs of all stakeholders and,
facilitate access to effective tools and support so that IATI data contributes to the achievement of sustainable development outcomes.
Since 2008, when IATI was launched at the Third High Level Forum on Aid Effectiveness in Accra, the IATI datastore has grown to include over 1784 development actors publishing information about their activities. And whilst it’s not perfect, it is already an amazing starting point as a resource for tackling the challenges to aid effectiveness described above.
The problem is that despite the wealth of data, development organisations have the problem that it takes too long to extract useful insights from that data and so it is usually not worth the opportunity cost of their staff's time. This is where DevExplorer comes in. DevExplorer uses the power of Large Language Models (LLMs) to reduce the cost and time to extract useful insights from the wealth of information in IATI. It creates a simple and easy to use AI powered workflow to search and create custom analysis based on the IATI dataset that could support each of the challenges described earlier (see previous blog for longer description of the workflow based approach).
Ownership of development priorities by developing countries: A developing country stakeholder could quickly create different analyses of the development activities taking place in their country.
Focus on results: A development actor interested in what results are being delivered or how to improve delivery by learning from previous projects could create analyses of results or lessons learned in a specific geography or sector.
Inclusive development partnerships: A development actor who wants to make collaborate and coordinate with other development actors can create a summary of what other development actors are doing in their area of interest.
Transparency and accountability to each other: Anyone who is interested in holding development actors to account can more easily get bespoke summaries of what development actors are doing.
The process of searching for, extracting, summarising and categorising key lessons, results or partnerships from 50 programmes and presenting it in a useful format would likely take a person over 20 hours. DevExplorer enables the same person to accomplish the same task in under 10 minutes. This doesn’t just save time, it means that people can actually do this analysis where previously they would never have considered doing it. And if analysis like this became common place it would have a significant effect on reducing the knowledge barrier to living up to the aid effectiveness principles.
DevExplorer is still in development. Currently, we have only focused on FCDO data and we are in the process of extending it to a broader range of IATI publishers, building in additional functionality based on user feedback and conducting an evaluation of the tool. Even then the tool will be only as good as the data available in IATI (or other development datasets). However, we see DevExplorer as just the beginning of an era of reimagination of the power of data combined with AI to improve aid effectiveness.
Publish date: 23rd June 2025
