Every breakthrough starts with an idea. From the theory of evolution to the discovery of DNA’s structure, pioneering research is at the heart of the University. In 2024 alone, the University secured more than £580 million in research income, helping to fund the next generation of discoveries.
But behind every brilliant research idea lies a mountain of paperwork. Identifying funding opportunities and developing proposals to help bring groundbreaking research projects to life often entails hours of painstaking administration.
“A major problem when applying for grants is that individuals often come across funding calls they’ve never, or rarely, seen before,” explains Petra Georgoulis Hluzova, Research Grants Manager for the Department of Psychology. “We need to quickly find all the relevant information that is often scattered across websites, webinars, or 150-page PDF documents. We have to be fast, and there is little room for error.”
The stakes are high. Missed opportunities or incorrect information potentially cost the University millions of pounds in lost revenue for research funding. The scale of the challenge is particularly stark in the Department of Engineering – the University’s largest department. The department is especially research active, submitting upwards of 600 applications each year. Here, a team of four research coordinators can spend upwards of five hours each every week hunting down answers to specific grant queries, according to the department’s Head of Research Support, Peter Thomas-McEewen.
Small improvements could make a big difference
The challenge sparked a collaboration between research support staff in professional services roles across the University to explore whether AI could streamline the research funding process – from summarising funding calls to checking eligibility and interpreting complex terms and conditions.
“With hundreds of applications submitted each year and millions of pounds of research income at stake,” explains Petra, “Even small improvements in how we work can make a significant difference.”
Alongside Petra and Peter, a team of research support colleagues from across the University – from science and technology through to arts and humanities – secured funding through ai@cam’s AI for Ops programme to test whether AI could reliably and accurately speed up elements of the grant application process.
The project began with a series of focus groups to gather a range of views – from established academics, early career researchers and other professional services staff – to identify common research queries across disciplines.
Rather than relying on a single AI platform, the team evaluated a range of leading tools, including ChatGPT Business, Microsoft Copilot, Google Gemini and NotebookLM. By testing each platform using the same prompts and queries, they were able to compare their accuracy and reliability against traditional manual searches. The results showed that AI could answer a range of queries, including those around eligibility criteria, much faster than traditional methods while achieving similar levels of reliability.
Empowering rather than replacing humans
The ultimate vision is to put AI tools at the “fingertips” of research support staff, allowing them to handle the “day-to-day churn of straightforward questions” and making sure they have more time for the complex, nuanced queries. Such resources would significantly reduce a very steep and lengthy learning curve for new staff and could also be used to enable academics and researchers to find some answers themselves, without lots of back-and-forth emails. The tentative hope is that AI could help reduce workloads by at least 20%.
“I’m really interested in the potential of AI for resolving a range of query types,” says Peter. “My team can then spend more time providing feedback to applicants on the proposal itself and supporting more applications. It also allows us to add value in other areas, like managing live grants and providing targeted support packages.”
The project team is clear that AI will be used as a tool to empower and train colleagues – enhancing rather than replacing human expertise.
“There are polarising views on AI,” acknowledges Peter. “One person in a focus group was worried about whether they would still have a job in a few years. We’ve been very clear: this is about adding efficiencies and enabling people to put their time into other activities. In no way are we suggesting anyone is being replaced. There will always be the need for people in these roles – bringing human decision-making to complex scenarios and facilitating a wide range of tasks within the application process.”
The team is acutely aware of other sensitivities around using AI, particularly data security and the reliability of information. To address this, the project established a dedicated compliance subgroup to ensure its work remains aligned with the University’s AI policies. The scope also relied exclusively on publicly available information from funder websites, steering away from any sensitive personal or financial data.
Sparking wider conversations
Although the initial focus was on research support, the team hopes that similar AI-driven approaches could be applied to other areas across the University where staff manage complex documentation – such as rules and regulations for successful grants, human resources, procurement and compliance.
None of this work would have been possible without AI for Ops funding, which has helped to bring the project to life, according to both Peter and Petra. “We wouldn’t necessarily have driven this project forward without this funding opportunity, especially for accessing more advanced licences to develop and share agents,” says Peter. “Crucially, it’s enabled us to have a more collaborative project with a bigger team, which is really helpful in terms of insights and ideas for the project.”
Beyond the technical outcomes, the project has already sparked valuable conversations and insights about the role of AI in research administration. “We hope we can demonstrate how AI can make a meaningful difference in this space,” says Petra. “We’ve shown the value of bringing together colleagues from across the University to explore new ideas collaboratively. That’s something worth building on.”