Co-Writing Grants with AI

By: Sarah Pita

If you use AI for a long, multi-part document like a grant application, be sure to take it question by question. It may seem efficient to upload the full application and ask the AI to complete it for you. But if you do it that way, the application will not benefit fully from your expertise as a grant writer, and the AI will not be doing its best work either.

Generative AI tools have limits on how much information they can handle at once, and if you dump in a whole grant application at once, it is less likely to provide the kind of detailed response needed for a winning application. The result is more likely to be repetitive and vague.

And you will be less engaged with each individual response if you’re reading the whole generated thing at once. AI does not err in the same ways humans do. Its mistakes tend to be structural rather than superficial. Big mistakes, not small ones. But if you are used to looking for grammar, run-on sentences, and awkward phrasing, you may miss those structural flaws altogether.

By contrast, if you go question by question, you have to engage more actively with each question, and with the response you want for each one. You can ensure that each question gets a complete answer—important in processes where each response is scored separately.

Use AI as a writing companion, but supervise it closely. After all, you know more about this project than it does.

Here’s how:

At the start of the project, explain what you’re trying to do (“We are applying for a general operating grant from the Wombat Trust”)  Share links, or upload:

·       Websites (yours and the funders)

·       Organization information like your Annual Report or relevant one-pagers

·       Current program information—even old grant applications.

·       The actual application form or RFP.

And then, BEFORE YOU HIT ENTER, say “any questions?”

Your AI of choice should have some questions. If it doesn’t, it might not understand the assignment. Reread your original prompt and see if there is anything unclear there.

Then say, “we are going to go through this application question by question. We’re starting with question 1.” Explain how you want to answer. In a way, your prompt here is a stealth first draft, with none of the fear of the blank page. Just tell the AI how you want it to answer, and anything it needs to emphasize. If there is a word or character limit, say that. Hit enter.

Thirty seconds later you’ll have a response.  READ IT CAREFULLY. You are looking for hallucinations (“Wait, we don’t have a music program!”), logic errors, and things that are missing either because you forgot to mention them or the LLM left them out.

Tell it what to fix and almost immediately you will have a revised response. Read this one carefully too.  

Don’t move on to the next question until you’re happy with the first one.

As you work, copy and paste the narratives into a Word document with the questions included.

At the end, read it over and make edits. Just like in the olden days.

Then, if you want, open a fresh chat (or, if you are feeling fancy, a different AI tool—have Claude critique ChatGPT or vice versa!) and ask it to look for inconsistencies, or compare it to the RFP again. But be firm: “no rewrites.”

This method is faster than writing a grant without AI, but I contend that you’re still writing a grant. You’re just letting the machine handle larger chunks. Kind of like when we switched from typewriters to word processing (for those who remember those days).

One big caveat: only do this if you are allowed. If the terms of the grant prohibit AI use, don’t use AI.

About the Author:

Sarah Pita is a fundraising professional with 25+ years of experience and a dynamic speaker who makes AI approachable and immediately useful for nonprofit teams. She leads practical, engaging trainings and workshops on using AI for fundraising and has presented at groups such as Women In Development NYC and at the AFP GPC Leading Philanthropy conference, among others. Sarah is currently Director of Development at the Center for Independence of the Disabled, New York.

Interested in an AI workshop or training? Contact Sarah here.