Research teams tackling global health challenges can now test artificial intelligence (AI) co-scientist tools on their own research questions, through guided sprints funded by the UK Foreign, Commonwealth & Development Office (FCDO) and delivered by PATH.

The sprints mark the latest milestone for FCDO’s AI for Development Science Breakthroughs (AI4DSB) programme. Running from October 2026 to June 2027, they give research teams hands-on, supported access to a new generation of AI tools designed to accelerate scientific discovery.

From launch to delivery

In June 2026, GRTD reported on the launch of AI4DSB. Announced at the Global Partnerships Conference, the programme set out an ambition to use AI to speed up scientific discovery and deliver tangible benefits for communities around the world. It builds on the UK government’s work through the AI for Development (AI4D) programme, with priority areas including drug discovery, crop engineering and weather prediction, and a commitment to making emerging AI-for-science tools more globally inclusive.

Since then, PATH’s Using AI to Accelerate Global Health Research project, the programme’s global health strand, has made strong progress. The team has already catalogued more than 15 AI co-scientist platforms and is building an open benchmarking framework to test how they perform on the tasks that matter in research, from synthesising published evidence to generating new hypotheses and designing candidate molecules. A living landscape map sets out what each tool aims to do, the evidence behind those claims, and what it takes to use them. The best-performing tools are already being applied to real global health challenges in antimicrobial resistance and malaria vector control.

This work fills an important gap. The field is moving fast, but there is not yet an independent, standardised way to test these tools. Without one, researchers risk chasing dead ends, funders risk backing tools that give convincing but wrong answers, and promising tools from smaller teams risk being overlooked.

The next step: guided sprints

The sprints take what PATH has learned and puts it in the hands of researchers. AI co-scientists go well beyond a general-purpose chatbot: they work through a scientific problem in repeated cycles, breaking it into sub-problems, synthesising large bodies of literature, generating and critiquing hypotheses, and suggesting ways to test them. Some also draw on specialised models, for example to predict protein structures or molecular properties.

Each sprint is a focused one-to-two-week collaboration built around a question the research team brings. PATH helps select and set up the most appropriate tool and helps run it against the question, and PATH and the research team interpret the outputs together. PATH brings expertise in getting the most from these tools, while the research team brings the domain knowledge to guide the AI and judge whether its outputs are sound. The programme covers access to the tools and PATH’s time; teams contribute their own staff time and the cost of any follow-on work they choose to pursue, such as laboratory validation. No prior experience with AI is needed.

The best candidates for these guided sprints are well-defined scientific questions where synthesising a large body of evidence, or generating new hypotheses, could genuinely move things forward. PATH’s own team is using co-scientist tools to identify biomarkers that could act as correlates of protection for infectious diseases, design new mosquito-killing compounds that work differently from existing insecticides and propose candidate molecules for low-cost point-of-care diagnostics. But sprints are not limited to molecular design. Teams might use a co-scientist to weigh up the evidence on targets for a new class of therapeutic, help design a molecular surveillance study across multiple sites, or create and troubleshoot a laboratory protocol.

“We’ve been testing co-scientist tools across several use cases, from designing new mosquito-killing compounds to proposing candidate molecules for low-cost diagnostics. They’re generally quite good at working through a large body of evidence quickly and at putting forward new and interesting ideas, but their outputs are only as useful as the expertise used to judge them. These sprints are designed to give research teams the chance to find out, on their own questions, where these tools help and where they don’t.”

Dr Rob Amato, Deputy Director of AI for Science, PATH

Every sprint is also a real-world test of how these tools perform. Lessons from the sprints, alongside the benchmarking results, will feed into a published learning agenda, so the wider research community can benefit, and AI-for-science tools can be put to work where they are needed most.

Find out more

The guided sprints are open to research teams working on global health challenges who have a defined scientific question that could benefit from AI-assisted hypothesis generation. Applicants are expected to commit staff time and include at least one domain expert who can critically assess the outputs, along with any data access and approvals their question requires. Teams working in or with partners in low- and middle-income countries are particularly encouraged to apply.

Applications are reviewed on a rolling basis, with sprints expected to run until June 2027. Where a question looks well suited to co-scientist support, PATH will invite the team to provide further details to complete the assessment and plan the sprint together.

If that sounds like your team, PATH would like to hear from you. Complete the Sprint Intake Form.