Exploring the Role of AI in Grant Applications

There has been a great deal of discussion in the grant-making world about the role of AI.

We foresee a world where AI-generated grant applications are being assessed by AI, so what is the role of the human?

At the Burdett Trust for Nursing we are experimenting.

For our recent Neighbourhood Health: Community Engagement Awards we introduced a toggle switch on our online application. Applicants had to choose between two options:

• This application has been prepared with the assistance of AI. Everything within the proposal is true and I have undertaken the final edit myself.
• This application has been completed without AI assistance.

We tried to make the decision value-neutral, so that people felt able to answer honestly.

AI provides a very useful tool for writing grant applications, particularly if the person crafting the text is new to the process or does not have English as a first language. How handy to have a first draft when you’re short of time! But we wanted people to declare whether they had used AI or not, so that we could see if there was any difference in the final submissions.

70 applications were submitted. 36 said they had used AI, and 34 said they had not. They were all assessed against clear scoring criteria by human assessors who were unaware of whether AI had been used or not. The panel was convened and funding decisions made.

Meanwhile, out of curiosity, we persuaded an (interestingly reluctant) MS Co-Pilot to assess each application using the same scoring criteria. It kept telling us that grant decisions should not be made using AI judgment (that’s good!), but when we explained this was an experiment, it agreed to engage. The result was a complete set of human and AI scores.

What did we find in our small-scale experiment?

• There was no significant difference in scores between applications that used AI and those that did not.
• Applications prepared using AI were scored slightly higher by CoPilot.
• Applications created without AI were scored slightly higher by human assessors, who did not know whether AI had been used.
• Human scoring ranked the applications very similarly to CoPilot’s ranking, but Co-Pilot scoring was consistently higher.

Please bear in mind the sample is too small for this to be significant, but it is interesting, nonetheless.

We think these findings demonstrate that the scoring framework was robust enough to be applied consistently by humans and AI. The pattern of higher AI scores is intriguing. My hunch is that, as nurses assessing nursing applications, we were looking for evidence of a deeper understanding of the issues asked for, and our scoring went beyond just using the right words.

I think these findings provide reassurance for applicants and reviewers alike.

In this sample, the use of AI tools does not appear to advantage or disadvantage applicants, but perhaps democratises the process, assisting those who are new to grant applications.

The ranking of applications was consistent between human assessors and Co-Pilot, so perhaps there is a case for continuing to experiment with AI-assisted sifting of applications. As we expand our use of AI across every area of our lives, the Burdett Trust for Nursing will continue to explore how emerging technologies enable our grant-making processes. This includes supporting applicants, maintaining transparency, ensuring fairness, and strengthening confidence in decision-making.

We’d love to hear your thoughts and experiences.

(PS This blog has been written without the assistance of AI!)