WFP's Enterprise Deduplication Service — AI Photo-Matching Saves $431,000 in Mali Aid Registrations

WFP's beneficiary databases, like many large humanitarian responses, risk duplicate registrations — the same person recorded more than once under slightly different personal details — which wastes scarce food-assistance budgets and slows the process of getting help to people who need it.

In 2025 WFP piloted its Enterprise Deduplication Service (EDS) in Mali, an AI tool built on open-source computer-vision models that compares beneficiary photographs, names and other registration details to flag likely duplicate records for human staff to review; humans, not the algorithm, make the final eligibility decision, and the design lets people keep religious or cultural coverings such as veils or turbans on for the photo match. The same tool has also been piloted in Afghanistan, Burkina Faso, Cameroon, Mozambique, Niger, Somalia and Uganda.

Over six months in 2025, EDS helped WFP's Mali operation save more than US$431,000 by identifying duplicated assistance, cutting a task that used to take staff weeks of manual spreadsheet comparison down to a matter of hours. WFP projects the tool could save at least US$4.7 million globally in 2026 as it scales, equivalent to roughly 6.7 million additional meals, and reports matching accuracy of 99.99%.

Because it runs on open-source AI models rather than licensed biometric software, WFP states it costs around 50% less than comparable proprietary deduplication systems. WFP says it runs data-privacy audits and assessments before each country deployment and stresses that AI flags are advisory only, with staff retaining sign-off; independent, published evaluation of false-positive and false-negative rates, and of any downstream effects on beneficiaries incorrectly flagged, has not yet been made public.