AI Tools for Grantmakers

The stack — and the guardrails — for funders and program officers who want AI to reduce toil without eroding the trust grantees have in them.

Where AI genuinely helps grantmakers

• LOI triage: summarising 200 letters of intent into structured criteria — never for auto-rejection.
• Due diligence synthesis: combining public financials, news, and prior reports into a briefing.
• Report reading: extracting outcomes from long narrative reports so program officers can spend time on the harder judgements.
• Board-pack drafting: turning your notes into a first-draft memo in your foundation's voice.
• Multilingual grant review: understanding proposals submitted in isiZulu, Sesotho or Afrikaans without waiting for translation.

Where AI must not be used

• Final funding decisions. A model does not carry your foundation's fiduciary duty.
• Ranking or scoring applicants against each other — bias in, bias out, with the appearance of objectivity.
• Communicating rejections. A human wrote the yes; a human writes the no.
• Pasting grantee financials or beneficiary data into a public chatbot.

A minimal, defensible tool stack

• Reasoning model: Claude Sonnet or GPT — for synthesis, summarisation, drafting.
• Enterprise workspace: a paid account with a zero-retention data policy, not the free consumer app.
• Private document Q&A: NotebookLM or a similar RAG tool where sensitive documents stay in your grip.
• Transcription: Whisper (or Otter with consent) for site-visit notes.
• A foundation voice brief: the highest-leverage artefact in your stack — see Module 2 · Description.

Publish your AI policy

Grantees are already wondering whether their proposal is being read by a model. Tell them. A one-page AI-use statement — what you use AI for, what you don't, what data never enters a model — is now table stakes for a trust-based funder.

Continue: AI in Fundraising · AI for Nonprofits South Africa