Can you use AI to write grant proposals?

I did not come to this as an AI enthusiast. I was an early adopter of ChatGPT in 2022, and it saved us time for about six months. Then something happened that nobody wants to say out loud: we did not do the same amount of work more efficiently. We just did more work, faster. That is not a win. That is a treadmill with a new motor.

What changed for me was reframing the question. Instead of asking AI to help me write, I started asking what administrative work was churning and burning my time, and whether it could be handed off entirely. That is a different question, and it has a much better answer.

• Grant research: finding candidates, verifying them, and scoring fit against your actual budget and eligibility.
• Pipeline building: turning a ranked list into a sortable sheet with funder, program, award range, deadline, fit score, deal breakers, and next action.
• Budget translation: recoding your annual operating budget into a funder's specific template, including the indirect cost percentage rules that eat an afternoon.
• First drafts: a complete proposal draft, a submission checklist, and notes on what is still missing from you.
• Context recovery: pulling the last meeting notes, the email thread, and the file you cannot find, so you are not spinning your wheels before you start.

I am here to teach you to outsource the admin, never your thinking. Your judgment about which funder is worth a relationship, what your program actually does, which numbers are true, and how to talk about the people you serve is the asset. It is the one thing that cannot be replaced, and it is the first thing people accidentally give away.

The fear I hear most is that AI will make us lazy writers. I used to have that fear too. What happened instead is that I taught it to write like an entry level grant writer, and now I edit with a fine tooth comb. The proposals are better, not worse, because my attention is on the argument instead of on the formatting.

There are two ways people use AI, and confusing them is why most nonprofits stall out after a month.

Prompting

You ask it to do something. Help me answer this question. Rewrite this so it matches the funder's priorities. Then you sit there and watch it spin. Useful, but you are still in the chair, and the work stops when you stop.

Architecting

You build a workflow that does the work without you in the chair. It knows when to start, it knows the whole sequence of tasks, and it files the output where it belongs. That is the difference between a tool you operate and an employee you trained.

Building a good prompt is a lot like training an entry level employee. If you told a new hire to go find ten grants we can apply for, you would get back a face full of questions. Where should I look? What do we qualify for? How big? By when? AI is no different. It just does not ask.

• Role. Tell it who it is. You are a grant research analyst for our organization.
• Context. What it needs to know about you. A 501(c)(3) in Northwest Arkansas that gets more people onto mountain bikes, with a focus on women and underrepresented riders, through events, skills programs, and community building.
• Task. Exactly what you want. Find ten grant opportunities we could realistically pursue in the next six months.
• Format. What the output should look like. For each one: funder name, program name, typical award size, deadlines, and one sentence on why it fits our mission.
• Constraints. The guardrails. Prioritize funders with a history of supporting outdoor recreation, community health, youth development, or women's sports. Do not include funders that only give to universities or hospitals.

That last piece is where most people leave value on the table. Constraints are how you stop wasting your own time reading through opportunities you were never eligible for.

A list of ten funders is not a pipeline. The next move is the one that changes everything: instead of clicking through each opportunity and applying your own judgment, you teach it your judgment once.

Score each opportunity from one to ten for fit. Use these criteria: mission alignment, award size relative to an annual budget of $250,000, whether we are clearly eligible, and how heavy the application is. Show one sentence of reasoning per score and flag any deal breakers.

Now you get back something a ranked list cannot give you. Strongest alignment to effort ratio on the list. The only thing standing between you and the 2027 cycle is a five minute interest form. Three of your top five require an action that is not an application. Do not build a revenue projection on these three until someone confirms award numbers by phone.

Ask it to put the ranked list into a spreadsheet with columns for funder, program, award range, deadline, fit score, deal breakers, and next action. Sort by fit score, highest first. Make the next action column specific, like confirm eligibility by phone or request LOI guidelines.

I have built that sheet by hand more times than I can count. It takes hours, and updating it takes hours again, because verifying eligibility and cycle dates for every line is the actual work. I do not build it by hand anymore. Neither should you.

For a live application, the input is three files: the blank application questions copied out of the funder's portal, the RFP, and the funder's budget template. Add an approved copy of the organization's annual budget, and approved means the organization knew and agreed to it going into AI.

What comes back is four things, filed in the right folder:

• A full proposal draft written to the actual questions.
• A budget recoded into the funder's template, inside every constraint, including the indirect cost rules with the nested percentage caps.
• A submission checklist.
• Notes on what it still needs from you.

The budget is the part that surprises people. If you have ever fought a grant where you can use thirty percent on indirect expenses but that amount cannot exceed some other figure, and you are hand building Excel formulas at eleven at night, that is the work that goes away first.

I am going to say fact, and tell you that it took mine. This company started as a grant writing and fundraising consultancy. My client base for that specific service is about ten percent of what it was in 2024.

That is not a reason to avoid this. It is the reason to learn it. If you write grants for a living, the skills in this article are a competitive advantage right now, and the window where they are an advantage rather than a baseline expectation is not going to stay open forever.

• Get organizational approval before uploading budgets, donor data, or anything with personal information in it.
• If your IT department manages your computer, get approval before installing anything.
• Verify every number and every deadline. AI is confident when it is wrong.
• A human signs off before submission. Every time.
• Be eyes wide open about the infrastructure. Data center construction has real environmental and community costs, and using these tools responsibly includes knowing that.