Project proposals and fundraising: AI as a drafting assistant

1Analysing a call for proposals

A call for proposals is often 30 to 80 pages long, with annexes. AI helps break it down quickly, provided you give it the complete document:

• "List all the eligibility criteria (organisation, partners, areas, amounts, duration) in a table, with the page of each criterion."
• "List all the documents to be provided and the required formats (number of pages, font, language, annexes)."
• "Summarise the evaluation grid and the weight of each criterion."
• "Which requirements are likely to be overlooked by a hurried applicant?"

Then check each element in the original text, especially the deadline, the time zone, the minimum and maximum amounts and the co-financing rules. A single eligibility error gets an application rejected, however good it is.

Go / no-go decision. Also ask: "Given our profile [presence, experience, annual budget, partners], which criteria do we meet, which are uncertain, which do we not meet?" The decision to apply remains management's.

2Theory of change and logframe

AI is useful for structuring the logic of a project from your needs analysis:

• proposing a theory of change (if... then... because...) with explicit assumptions;
• proposing a hierarchy of objectives (overall objective, specific objective, results, activities);
• suggesting SMART indicators and their means of verification;
• checking consistency: does each activity contribute to a result? Does each result have an indicator?

Points of vigilance:

• Numerical targets and baselines come from your data and your real capacity, never from AI.
• Prefer the standard indicators required or recommended (Sphere standards, cluster indicators, donor indicators): give the list to the assistant.
• Assumptions and risks must reflect your real context: security, rainy season, access, prices, authorities.

Standard request. "Here is our needs analysis and the list of cluster indicators. Propose a logframe in a table. Do not invent any target: leave the target column empty. For each indicator, state whether it comes from the cluster list or is a proposal."

3The narrative: from first draft to the organisation's text

For the narrative sections (context, justification, approach, sustainability, gender, accountability, localisation), an effective method:

• Your elements first: list in bullet points the facts, experience, partners, lessons learned. AI does not know your added value.
• First draft by the AI, section by section, within the page limit.
• Rewriting by the team: adding concrete examples, verified figures, place names, real lessons. This is what distinguishes a winning proposal.
• Critique by the AI: "Assess this section against the call's grid [grid]; score each criterion out of 5 and propose improvements."

Evaluators read hundreds of applications. Empty phrases ("holistic and innovative approach", "multi-sector synergies") churned out by AI lower the score. Concrete, local and verifiable content raises it.

4The budget and its justification

The budget remains specialist work, using the organisation's financial tools. AI can help to:

• write the budget justification (narrative) from the detailed budget;
• check consistency between narrative activities and budget lines ("List the narrative activities without a budget line, and the lines without a corresponding activity.");
• explain or write spreadsheet formulas for the calculations.

Never trust it for final calculations without checking: recalculate in the spreadsheet. And do not give it named salary information.

Control example. The narrative announces 12 training sessions of 25 participants, but the budget provides per diems for 200 participants. 12 × 25 = 300 participant-sessions; if each person attends 3 sessions, that makes 100 people, not 200. The inconsistency must be corrected before submission: this is exactly what AI can help spot, and what the team must settle.

5Donor rules on AI

Donors and foundations have begun to regulate the use of AI in applications and reports. Positions vary: some accept its use as a drafting aid, others require it to be disclosed, others recall that the applicant remains fully responsible for accuracy and originality. Several research funding bodies also prohibit reviewers from using AI on the confidential applications they receive.

Good practice:

• read the terms of the call and the donor's guidelines on AI, if any;
• disclose use when required, factually;
• do not give a consumer AI documents that the donor has marked confidential;
• keep track of your sources and versions;
• check that the proposal promises nothing the organisation cannot deliver: it is a contractual commitment.

6Beyond proposals: the donor relationship

AI also helps with the rest of fundraising:

• Monitoring funding opportunities with web search and sources (then checking on the donor's official website);
• Donor profiles from public documents: priorities, countries, amounts, procedures;
• Short concept notes from a long proposal;
• Meeting preparation: "What difficult questions could this donor ask about our exit strategy? Propose elements of response from our document.";
• Messages to private donors: thank-you letters, newsletters, in line with communication ethics.

The relationship of trust, however, rests on people: visits, transparency about difficulties, accurate and timely reports.