Using AI assistants for proposal drafting, safely
How to use an AI assistant to help draft donor proposals without risking confidential data, invented facts or a generic voice: a safe, step-by-step workflow.
Short answer
Use an AI assistant for proposal drafting only as a helper: give it your own facts, ask for a structure or a first draft of one section, and never paste confidential or personal data. Check every number, name and claim, rewrite in your organisation's voice, and follow the donor's rules on AI use. You remain responsible for everything you submit.
Key points
- Give the AI your facts; do not let it supply facts.
- Never paste personal data, security details or confidential documents.
- Use AI for structure, plain-language edits and first drafts of single sections.
- Check every figure, name, date and source by hand.
- Read the donor's guidance on AI use before you submit.
Can AI help write a proposal?
Proposal deadlines are tight, and many teams write in a second language. An AI assistant can help with structure, clearer wording and first drafts. OCHA's briefing note on AI in the humanitarian sector says AI can improve efficiency in routine tasks like report drafting. It also warns of misinformation, bias and erosion of privacy. A proposal sits between the two: it is a drafting task, but it carries your organisation's name, data and promises.
The safe rule is simple. AI helps you write. It does not decide what is true, what the community needs or what you can deliver.
What should you never put into an AI tool?
Public AI tools may store what you type. Check the provider's terms. Keep these out:
- Names, photos or contact details of beneficiaries or staff.
- Locations or details that could put people at risk.
- Confidential donor documents, partner agreements or unpublished budgets.
- Login details and bank information.
Use general descriptions instead. Write "a township in a conflict-affected area" instead of a village name. For a wider look at the risks, read how NGOs can use AI tools responsibly.
A safe workflow in six steps
- Read the call for proposals first. Note the format, word limits, criteria and any donor rule on AI.
- Build your facts file. Write the problem, target group, activities, results and budget notes in your own words, with sources. This is your material.
- Ask for structure. Paste only the call's headings and ask for a section outline. Compare it with the donor's template and keep what fits.
- Draft one section at a time. Give the AI your facts for that section and ask for a plain-language draft. Say what to avoid, such as invented statistics.
- Check and rewrite. Verify every number and claim against your facts file. Replace generic phrases with specific local detail. Keep a short note of which parts used AI help and who checked them, in case the donor or your director asks. Our proposal writing checklist shows what reviewers look for.
- Get a human review. A colleague who did not use the AI reads the full proposal for logic, voice and accuracy.
What are good and poor uses?
The table below compares typical uses. The pattern is clear: AI is good at shaping words you already have, and poor at supplying evidence, judgement or local knowledge that your team must provide.
| Good use | Poor use |
|---|---|
| Turn rough notes into clear sentences | Ask it to "write the needs section" with no data |
| Shorten a section to fit a word limit | Ask for statistics to support your case |
| Suggest headings for a concept note | Paste a confidential partner budget |
| Check grammar and consistency of terms | Submit the draft without reading it |
| Draft a plain-language summary of your results framework | Let it design your indicators alone |
How do you keep your own voice?
A proposal is a promise, and reviewers want to hear a real organisation behind it.
Generic text is easy to spot, and reviewers read many proposals. After the AI draft, add what only your team knows: the history of your work in the area, what communities told you, a real challenge and how you handled it. Cut filler words. Read it aloud. If it sounds like any organisation, rewrite it until it sounds like yours.
How do you check for invented facts?
AI can produce fake figures, fake references and wrong dates. For each factual claim, do three things: find the source, check the number matches, and record the source in your notes. If you cannot find a source, delete the claim. Do the same for donor policy statements, because AI may quote rules that do not exist.
What about donor rules and honesty?
Donors differ. Some state a position on AI in the call; others are silent. Follow what the call says, and if you are unsure, ask the donor's contact before you submit. Never claim work, partners or results that the AI added and your team did not do. Remember, you sign the proposal and you answer for it. Our guide on writing good AI prompts for NGO work shows how to ask for what you need.
How EMPO can help
EMPO is a reliable, Myanmar-based training and consultancy provider and one of the country's leading names in practical NGO training, with 1,200+ people trained and 15 partner organisations. We offer proposal writing support and training, online or in person across Myanmar, in English and Burmese. See our services or contact us to discuss your next proposal.
Sources
- Briefing Note on Artificial Intelligence and the Humanitarian Sector — OCHA, 2024
- Data Responsibility — OCHA Centre for Humanitarian Data
Questions people ask
Will donors reject a proposal written with AI?
It depends on the donor. Some have rules on AI use and some do not. Read the call for proposals, and ask the donor if it is unclear. Weak, generic content is a bigger risk than AI itself.
Can AI write the whole proposal?
It should not. A proposal needs local knowledge, real data, a realistic budget and your organisation's own voice. AI can help with parts of the writing, but your team must design the project.
Can AI help with the budget?
It can help check arithmetic or suggest budget notes. Build the budget yourself from real costs, and confirm all figures against donor rules and your finance team.