Can AI write grant applications for a nonprofit?
The writing is the easy part, and it is not where the value is. Marlowe does the research, the eligibility screening and the compliance checking that decide whether a submission is worth making at all. It cut a modeled year-one development program from 981 hours to 297, found four funders that appear in no grant database, and its first significant act was telling the organisation to stop soliciting until it was legally cleared to ask.
The situation
A startup nonprofit building a trades high school. No development staff, no grant history, no track record for a funder to point at. Grant funding stood between the mission and the doors opening, and hiring a development office was out of reach.
This is the classic cold-start problem in nonprofit fundraising. The famous foundations reject organisations without a track record almost automatically. The funders who do back startups are small family foundations that do not run websites, do not publish RFPs, and do not appear in the databases you would pay to search.
What we built
Marlowe is configured like a development office rather than a writing tool. Four components:
- A research protocol — a defined method for finding funders, not a keyword search.
- A scoring rubric — explicit criteria for what makes a prospect worth pursuing for this organisation.
- A compliance guardrail layer — legal eligibility gates checked before a dollar of effort is spent.
- Institutional memory — a persistent record so the same ground is not covered twice.
The research method is the interesting part. Instead of listing famous funders that reject startups, Marlowe reverse-engineered a peer organisation’s public tax filings — the Form 990s that show exactly who actually gave that organisation money. Four of the nine funders it scored cannot be found in any grant database. They are only visible in someone else’s filed paperwork.
What it produced
| Result | What it was |
|---|---|
| ~70% less | Human labor for a year-one development program — a modeled 981 hours down to 297. |
| ~4.5× | More grant submissions possible from the same founder hours. |
| 4 of 9 | Scored funders that appear in no grant database, found in public tax filings. |
| 3 | Compliance exposures caught before any became a liability. |
| $32k–$84k/yr | Modeled cost of the human equivalent. |
The compliance catch deserves its own sentence. Marlowe stopped the organisation from soliciting at all until it was legally cleared to ask. Charitable solicitation registration is state-by-state, unglamorous, and routinely missed by new organisations — and getting it wrong is the kind of problem that follows an organisation for years.
Straight talk
No grants have been won yet. The organisation has not submitted. We are not going to imply otherwise by leaving that out.
What changed is what is possible per founder-hour — and the honest headline is that its first big save was saying “stop” at the right moment, not writing a beautiful narrative.
A human still writes the relationship. Marlowe drafts, screens and researches. It does not take the meeting, and family foundations give to people they have met.
The hour and dollar figures are modeled against published rates for development staff, not an audited comparison.
Where this pattern fits
Startup nonprofits with no development staff. Small organisations where the executive director is also the fundraiser. Any organisation whose funding sources are public record but scattered — which describes most philanthropic and public funding.
The same structure works outside the nonprofit world wherever eligibility screening is the expensive part: grant programs, incentive schemes, certification-gated markets.
It fits badly if you have a strong development director and a warm funder network. Then the bottleneck is not research, it is time in front of people.
Bring the funding work nobody has hours for. The audit maps your operation, ranks the top five opportunities by return, and gives you a written plan — whether or not you build with us.
Book Your AI Opportunity Audit — $1,000, credited toward your build