01 The Problem
A non-profit’s conservation team depended on grants but spent huge amounts of time hand-writing proposals customized to each funder, slowing them down enough to miss deadlines.
02 The Approach
Proposal Structure
Mapped out the 5 sections every strong proposal needs.
Prompt Design
Built one structured prompt that has GPT-4 fill all 5 sections from just a few inputs (name, objectives, timeline, budget, focus area).
Prototype Build
Built a working Python tool using the OpenAI API so staff could get a full draft from one input.
Testing
Ran it on a real example project to check the output made sense and had real numbers, not filler.
Pitch Deck
Packaged it into a slide deck for the department and board.
03 The Result
Full formatted proposal generated in under 2 minutes
From a short set of inputs to a complete draft.
Specific goals, justified budgets, funder-appropriate tone
Held up across repeat tests — real numbers, not filler.
Estimated 80% cut in drafting time
Hours of writing compressed into a short generation loop.
University grant writers reviewed the output and liked it
Outside review helped validate the prototype wasn’t just flashy.
04 Business Impact
Time Back
Frees staff for actual conservation work.
No Consultants
Removes need to hire outside grant writers.
More Funding Shots
Faster process means more grants applied for — estimated up to 50% more funding.
Consistency
Quality doesn’t depend on one staffer’s bandwidth that week.
05 What I Learned
Output quality lived and died on how specific my prompt was. A vague prompt gave a vague proposal.
06 What’s Next
Give it a real interface so non-technical staff can use it without touching code.