In May 2023, Rwanda experienced severe flooding that caused an estimated US$187 million in direct damage. The event became a case study in why timing matters: forecasts and early warnings are useful only if they reach farmers and planners early enough to inform their decisions. Rwanda’s feasibility study used the event to demonstrate what AI-powered nowcasting could have delivered differently.

On 3 September 2026, together with Meteo Rwanda, we presented the findings of a joint feasibility study on integrating artificial intelligence and machine learning into weather forecasting and climate services. The session brought together experts from Oxford, the United Arab Emirates, Ghana and the Gates Foundation, alongside Rwandan government officials and farmers.

The study’s central finding is economic, not technical: every dollar invested in this system generates a return six to seven times greater. Coffee alone, as a single value chain, stands to save US$1.4 million annually through better-timed disease control and gain between US$2 million and US$3 million in additional revenue by avoiding quality downgrading. Comparable early warning systems have generated returns of 16 to 1.

The investment case is no longer in question. The constraint lies in mobilising the necessary financing and aligning the ecosystem to use it effectively.

During the session, the Director General of the Rwanda Meteorology Agency offered a simple rule: AI is a good cook, but a good cook still needs ingredients. Without data, there is nothing with which to forecast. However, the insight goes deeper. Rwanda’s sequencing reveals what must happen in practice: first, a data foundation; then, people and institutions; and finally, the technology.

Most of Rwanda’s weather observation network is still manual. Real-time data are a prerequisite for early warning, so the network is being converted into a network of automatic stations. Observations from ground-based, radar, satellite and upper-air sources have not yet been integrated into a single system. Computing resources for training and testing forecasting models are insufficient. Dependence on foreign expertise also undermines sustainability. Rwanda is building the capacity of its own forecasters to implement and manage these systems; this is what makes the resulting products sustainable.

However, even that is not enough. The conversation revealed a systems gap that no single institution can close alone.

Professor Michael Obersteiner, who developed the benefit methodology underpinning the study, made a stark observation: a forecast has no value if it arrives after the decision has been made. A farmer whose seed is washed away planted three weeks earlier, but weather models are not reliable at the three-week horizon. The answer, therefore, does not lie in a better forecast alone. He proposed anticipatory finance, with payouts triggered by a forecast before a disaster occurs, so that a farmer can buy seed and replant immediately.

Dr Obai Khalifa of the Gates Foundation added another dimension: the test of an innovation is not whether it is adopted, but whether it changes a decision. Applications that diagnose crop diseases from photographs have existed for years but remain largely unused because nothing in the decision-making chain depends on them. What emerged was a systems question. The Government invests in core data and digital public infrastructure, but the ecosystem that uses that infrastructure includes insurers, input suppliers, aggregators and banks. Each stands to benefit from better predictions in relation to underwriting, investment and repayment terms. Yet none of them was in the room. That is the gap.

Rwanda has the advantage of a small, aligned government that is willing to move from dialogue to implementation. The Ministry of Environment, Meteo Rwanda, the Ministry of Agriculture and the Minister of State at MINAGRI are working from a common plan. However, Rwanda also has something less visible and more important: proof that the last-mile model works. The Green Trust Project, working with Meteo Rwanda, installed automatic weather stations at the community level and integrated a course on weather and climate services into the farmer field school curriculum. Farmers can now interpret seasonal, monthly, weekly, 10-day and five-day forecasts, resulting in measurable productivity gains. This is not a research project. It is institutional practice. This is why other African governments are watching and why the commitment made during the session was to build a pan-African coalition on climate intelligence, using Rwanda as a reference case.

The Government of Rwanda has identified US$3.5 million as the investment required to move from a blueprint to an operational national climate intelligence system. The investment roadmap is phased and de-risked. The economic case is compelling. The partners are aligned. What remains is to mobilise the necessary financing and, critically, convene the ecosystem actors, including insurers, input suppliers, aggregators and banks, that will use the information to inform decisions, rather than simply provide data.

The Minister of State at MINAGRI was clear about the stakes: do we continue spending billions to recover from climate-related disasters, or do we invest a fraction of that amount to anticipate them? The next El Niño is coming. Rwanda will face it better prepared than it was in May 2023, but only if it decides now to fund and align the ecosystem, rather than waiting until after the next shock occurs.

Rwanda now has the science. What it needs is the ecosystem.