From access to adoption: Case studies of AI adoption in science institutions

How are research institutions turning advances in AI into scientific progress, and what can their experience tell us about the next phase of AI for science policy?

Published by ai@cam in partnership with the Bennett School of Public Policy, From access to adoption examines ten initiatives across the UK and internationally to understand how research institutions are responding to the challenge of supporting safe and effective deployment of AI in science.

The case studies illustrate different approaches: from KU Leuven’s shared engineering service, which helps researchers apply AI to scientific problems, to AIchemy’s flexible funding for emerging research ideas; from Data Science Africa’s networks for developing skills and expertise, to ECMWF’s work testing and validating AI forecasting systems for operational use.

Together, they highlight the importance of what the report calls the ’missing middle’ in AI for science: the institutional capabilities needed to support experimentation, connect researchers with technical expertise, prepare scientific data, assess new methods, and sustain successful applications.

The report argues that AI adoption is a source of innovation. As researchers apply AI to scientific problems, they identify new opportunities and technical limitations that can inform the development of AI technologies.

These findings raise important questions for policymakers, funders, and research institutions about how to organise investment, sustain essential expertise and infrastructure, and ensure that the UK’s scientific strengths contribute to the future development of AI.

Download the report: From access to adoption: Case studies of AI adoption in science institutions (PDF)

Published 23 November 2026

ai@cam, University of Cambridge, in partnership with the Bennett School of Public Policy