What Makes a Strong AI-for-Social-Impact Project Proposal

A good idea isn't a good proposal. The difference is structure: a proposal makes your thinking legible to someone who wasn't in your head. Whether you're entering a contest or pitching a teacher or funder, strong proposals share the same backbone.

1. A specific problem, not a theme

"AI for education" is a theme. "Students in my town can't get past-paper practice in their language" is a problem. Name the who, the where, and the pain. Specific beats grand every time — a sharply defined problem is easier to solve and easier to evaluate.

2. A real user

Who, exactly, uses this — and have you spoken to one? Proposals that reference a real conversation ("I asked three nurses at a clinic and they said…") are instantly more credible than ones built on assumptions.

3. An honest role for AI

State plainly what the AI does and what it doesn't. "A model suggests a draft; a human reviews and decides" is stronger than implying the system is autonomous. If a simpler, non-AI approach would work better, say so — judges respect that judgment.

4. Ethics considered up front

This is where many proposals are thin and where you can stand out. Address:

• Bias — who might the data or model disadvantage?
• Privacy — whose data, and did they consent?
• Human oversight — where does a person stay in the loop?

These map directly to the UNESCO Recommendation on the Ethics of AI; referencing it shows you've done the homework.

5. Evidence and a measure of success

Cite a source for the problem's scale (a report, a statistic with a link). Then state how you'd know the project worked — even a simple, honest metric beats a vague claim of "impact."

6. Clarity over polish

A clear, plain proposal beats a buzzword-heavy one. If a non-expert can read it and understand the problem, the user, and your solution in two minutes, you've done it right.

A quick checklist

• Specific problem, named user
• Talked to at least one real person
• Honest, bounded role for AI
• Bias / privacy / oversight addressed
• A cited statistic and a success measure
• Readable by a non-expert

Where to build yours

Need ideas first? See 15 AI-for-the-SDGs project ideas. When you're ready to build and submit, the AI for Social Impact Challenge walks you through the proposal step by step and ends in a UNITAR certificate — read the contest rules for exactly how submissions are judged.