Grant Writing & Proposal Development with AI

Grant writing is the single highest-leverage task where AI can transform a nonprofit's capacity. A strong development director can expect to draft 2–3x more proposals per quarter with AI, without sacrificing quality — often improving it. This lesson walks you through the full AI-assisted grant writing workflow, from funder research to final polish.

What You'll Learn

• How to use AI to research and shortlist funders efficiently
• A repeatable workflow for drafting LOIs, needs statements, and full proposals
• How to use AI to align your proposal to a funder's evaluation criteria
• Pitfalls to avoid, including hallucinated statistics and compliance risks

Step 1: Funder Research

Before drafting anything, use AI to build a shortlist of funders. Perplexity is ideal here because it cites sources.

Sample Perplexity prompt:

List 12 private foundations that have awarded grants between $25,000 and $100,000 for {youth mental health programs in the Midwest United States} in the last 3 years. For each, provide: foundation name, recent grantees, focus areas, geographic scope, typical grant size, application process (LOI or full proposal), and a link to their grants page.

Then verify each finding against the foundation's own website or through Candid's Foundation Directory before spending time on an application. Never submit a proposal based on AI output alone — hallucinated funder details happen.

Step 2: Deeply Read the RFP

Once you have chosen a funder and retrieved their RFP or Notice of Funding Opportunity, paste the full document into Claude and ask a structured series of questions.

Prompt:

Below is the full RFP from {funder}. Produce (a) a plain-language summary in 250 words, (b) all eligibility requirements as a checklist, (c) the page and word limits for each section, (d) required attachments, (e) evaluation criteria with weighting if provided, (f) submission deadline and method, and (g) any disqualifying factors I must absolutely not miss. RFP: {paste}.

This single prompt eliminates the most common cause of rejected proposals — missed compliance details — and gives you a scaffold for the rest of the work.

Step 3: Build the Skeleton

Using the RFP summary, ask AI to build the full proposal outline with word counts.

Using the RFP details above and the following program summary, build a complete proposal outline for a {full proposal / LOI} to {funder}. For each section, include: section title, suggested word count, 3 bullet points of what to cover, and which evaluation criterion it addresses. Program summary: {paste}.

You now have a map of the entire proposal before writing a single narrative word.

Step 4: Draft Section by Section

Draft one section at a time. Fighting a blank page is easier in 400-word chunks than 4,000-word chunks.

Example — needs statement prompt:

Act as a senior grant writer. Draft the needs statement for the proposal outline above. Length: {word count}. Include 2 national or state statistics on the problem, one local statistic or data point relevant to our service area ({city/state}), and a one-sentence beneficiary description. Tone: urgent but not alarmist. End with a 2-sentence bridge to our program as the solution.

Repeat for each section — theory of change, program description, logic model, evaluation plan, budget narrative, organizational capacity.

Important: AI-generated statistics should always be verified. Claude and ChatGPT both occasionally invent plausible-sounding numbers. Run every stat through the original source before submitting.

Step 5: Align to the Funder's Evaluation Criteria

Once you have a full draft, run this alignment check:

Here is my full draft proposal and the funder's evaluation criteria. For each criterion, rate my proposal 1–5 and explain why. Then suggest 3 specific edits that would raise the lowest-scoring criteria by at least 1 point.

This is where AI earns its keep. Few development directors have the time to self-review with this rigor. AI does it in 60 seconds.

Step 6: Tone, Clarity, and Final Polish

Two final-pass prompts:

Review my draft for any jargon, passive voice, or redundant phrases. Flag each instance and suggest a clearer alternative.

Rewrite the opening paragraph three ways: one that opens with a beneficiary story, one that opens with a statistic, and one that opens with a bold mission statement.

Pick the best opening, and you are done.

A Real Example

A small literacy nonprofit in Texas used this workflow to submit a $75,000 proposal to a regional family foundation. Before AI: the development director estimated 22 hours of work over two weeks. With the workflow above: 8 hours over 4 days. The proposal was funded — the program director's feedback was that the final draft was tighter and better aligned to the funder's rubric than any previous submission.

Pitfalls to Avoid

• Hallucinated statistics. Always verify numbers against the original source.
• Invented funder history. Double-check any funder "grant history" the AI produces.
• Voice drift. If your organization has a distinctive voice, feed the AI a prior successful proposal as a style reference.
• Compliance blind spots. AI can miss page limits or font requirements. Always re-read the RFP checklist from Step 2 before submitting.
• Over-polishing. Funders can tell when a proposal reads as AI-generated boilerplate. Your final pass should add specificity: real names of neighborhoods, specific program sites, actual partner organizations.

Customizing for Federal Grants

Federal grants (HHS, DOJ, Education) are more complex than foundation grants. AI remains useful but with caveats:

• Break the process into smaller chunks. Draft one SF-424 narrative subsection at a time.
• Feed the AI the federal agency's scoring rubric before drafting. Federal grants are scored on points.
• Always comply with the word count exactly. AI will sometimes exceed limits — do a manual word count before submission.

Key Takeaways

• Use Perplexity for funder research, Claude for long RFP analysis, ChatGPT or Claude for drafting
• A repeatable workflow — research, RFP analysis, skeleton, section-by-section draft, alignment check, polish — can double your grant output
• Always verify AI-generated statistics, funder histories, and compliance details against the source
• Run your final draft against the funder's evaluation criteria with the AI as a reviewer
• The best proposals combine AI speed with human specificity: real names, real places, real beneficiaries