AI in funding: an experiment in time-saving

Our work for the Martingale Foundation was based on the Design Council’s Double Diamond methodology, which explains a project in four stages: discover, define, develop, deliver.

How we started

We started with a pro-bono project. We were lucky to be approached by a Laidlaw Scholar looking for an internship to develop their ethical leadership skills, contribute positively to society, and work on a project aligned with the UN Sustainable Development Goals. We found the Martingale Foundation as a result of a call on LinkedIn for an organisation looking for help where an AI might be a solution. Their time was invaluable to get us experimental tools that could show an impact. Our intern blogs about their experience here.

What we did

For that first stage of the project, we started with a workshop to discover some pinch points in Martingale’s process, and then defined solutions to two where we felt a short project could do some good. Our intern was mentored by a senior developer, and between them they developed two tools to alleviate specific sources of wasted time. One takes letters from student finance bodies submitted by students, and extracts a key datapoint. The other helps schedule the dozens of interviews Martingale needs to run.

We tested the tools with Martingale staff, to see if the help they gave was of use. It was; and that success secured project funding to re-work the tools for use on live data, helping save time on this year’s application process.

The code the project used is openly available here. It was written for Martingale’s specific situation, but the techniques we used will be of wider interest. Certainly, we’re glad to discuss it with other organisations in the third sector.

Result

Neontribe delivered two digital tools. One uses Python code to match candidates’ availability and research interests with academic assessors’ availability and research expertise for in-person interviews. The Foundation has found this tool saves days of staff time in scheduling interviews, vital in a tight admissions window.

The second uses GenAI to extract key data from PDFs into a readable field in their CRM, eliminating the need for staff to manually review hundreds of documents. This time-saving use of LLMs for document categorisation and targeted data pulls has clear potential for wider adoption across grant-giving charities as part of their due diligence processes.