How to use AI to draft a concept note without losing your voice

AI assistants can take a lot of the grind out of proposal writing. But reviewers read hundreds of submissions, and generic AI text stands out: vague claims, no local detail, and the same phrases in every paragraph. The trick is to give the AI your facts and your argument, and use it for structure, speed and polish.

This workflow works with any of the major assistants: ChatGPT, Claude, Gemini or Copilot.

Step 1: Summarise the call

Paste the call for proposals (or upload the PDF) and ask:

Summarise this call for proposals in a table: objectives, priority groups,
priority locations, eligibility requirements, maximum grant size, duration,
page limits, required sections, evaluation criteria with weights, and deadline.
Quote the exact wording for eligibility and evaluation criteria.

Check the output against the original. Deadlines and eligibility are the details AI is most likely to get subtly wrong, and the most costly to get wrong.

Step 2: Give it your facts

Write rough notes yourself: the problem as you see it in your district, your approach, your past results, your numbers and their sources. Bullet points are fine. Then:

Here are my notes for a concept note responding to the call above.
Using ONLY the facts in my notes, draft the "Problem" section in under 200 words.
Do not add any statistics, names or claims that aren't in my notes.
If something important is missing, list it as a question at the end instead of inventing it.

The last two lines matter most. They stop the AI filling gaps with invented figures and tell you what you still need to find.

Step 3: Draft section by section

Work through one section at a time rather than asking for a whole concept note at once. You'll get tighter text and it's easier to check. For each section, remind it of the word limit and the relevant evaluation criterion.

Step 4: Ask it to review like a donor

This is often the most valuable step:

Act as a reviewer for this call. Score my concept note against each evaluation
criterion out of 10 and explain the score. Then list the five changes that would
most improve it. Be critical.

You don't have to accept every suggestion, but it will reliably catch missing links between problem, solution and results.

Step 5: Tighten and fit the limit

Shorten this section to 150 words without losing any facts or numbers.
Use plain English and active voice. Keep the donor's terminology from the call.

Step 6: Make it sound like you

Read the full draft aloud. Remove phrases that sound generic ("in today's rapidly changing world", "holistic approach", "leverage synergies"). Add the concrete details only your team knows: a village name, what a community leader said, what didn't work last time and what you changed.

What to check before submitting

Every number has a source you can name.
No invented partners, past projects or results.
The donor's name and call title are correct throughout.
The page limit and format are met.
A colleague who knows the project has read it.

Is it acceptable to use AI for proposals?

Most donors don't prohibit it, but some now ask applicants to disclose AI use or confirm the content is their own. Read the call's instructions. Either way, you're responsible for every claim in the submission, and for keeping confidential or personal information out of AI tools you haven't checked.

For a wider view of tools for drafting, translation and data, see AI tools that save NGO teams time.