ONE launches an AI observatory for the world’s aid and debt data

ONE Data has opened the Development Finance Observatory, a free AI platform for finding and analysing the money flowing through developing economies. Launched on 14 October during the World Bank–IMF Annual Meetings, it brings together aid, debt and domestic-budget data from institutions whose figures usually live in separate databases. The ONE Campaign initiative was backed by $4 million in combined funding announced in January by ONE Data, Google.org and the Rockefeller Foundation.[1]

The attraction is a shorter journey from a question to a usable set of numbers. A journalist investigating debt payments or an organisation researching health funding can begin by describing what it needs, then obtain the underlying data. We tried that journey with a question about Mozambique.[2]

Five years, one sentence

On 18 October, we sent the public API this request: “Public and publicly guaranteed external debt service paid by Mozambique from 2020 to 2024, in current US dollars”. The service selected a World Bank International Debt Statistics series and returned five annual observations. We downloaded the DFO extract as a CSV bundle and compared its values with the original World Bank series, DT.TDS.DPPG.CD. All five matched exactly. This was a check of one query and one series; we did not test the signed-in analysis workspace.

The definition gives those figures their meaning. Debt service includes principal repayments and interest actually paid on long-term public obligations and on long-term private obligations carrying a public guarantee. The 2024 payment was about 109% higher than the 2020 payment, calculated from the unrounded figures. Both amounts are in the dollars of their respective years, without an inflation adjustment.[4][3]

What the AI has to understand

Before retrieving anything, the observatory turns the question into a precise description of the data: which measure, which country, which years and which units. Its knowledge graph gives countries and financial concepts shared identifiers, so observations from different sources can sit alongside one another. The number retains its source, date and unit as it travels through the system.[2][5]

This matters because apparently small words change a financial question. An aid commitment records an obligation; a disbursement records money transferred. A gross flow and a flow after repayments are different quantities. The observatory represents these distinctions explicitly. Its job includes translating the everyday question into the particular statistic that answers it.[5]

The same country identifier can connect a debt figure with other data about that country. That is the useful foundation for a more ambitious question: how do repayments, government resources and spending needs fit together? The first retrieval gives an analyst material to work with; explaining the relationship still requires choosing suitable comparisons.[5]

Try a bounded question

Get Data is available without an account. Start with a country, a clearly named measure and a fixed period, as we did with Mozambique. The steps below describe the documented website workflow; our hands-on check used its API.[6][2]

1. Open Get Data at developmentfinance.org and type your question. Include the currency and whether you want current or inflation-adjusted amounts when that distinction matters.
2. Read the sentence the service builds from your request. Its measure, place, years and source can be selected and changed. “About this answer” explains the interpretation and source choice.
3. Inspect the preview, then choose Download CSV. The download is a ZIP containing the data, source information, citation details and metadata. Other formats include Excel and Parquet.

For a report or briefing, choose a Snapshot when sharing the result: it preserves the captured rows. A Live Data link reruns the query when opened, so its figures can change with later releases. If a request is too broad to capture, narrow the countries, period or measure and try again.[7]

Mozambique’s five rows suggest the next question immediately: which creditors received those payments, and what drove the increase? A tool that gets a reader to that question—with the first set of figures already in hand—has done something useful.

Report an error or suggest a correction