Generate impact report narrative from data

The problem

You've got the data for your impact report: 127 people supported, 78% showed improvement, 45 moved into employment. But turning those numbers into readable narrative takes hours. You stare at a blank page trying to make statistics into a story. Funders want compelling narrative backed by data, and you're spending days writing reports.

The solution

Use an LLM to draft report narrative sections from your data and quotes. Give it the numbers, context about your programme, and sample beneficiary quotes, and ask it to write readable narrative. You then edit for accuracy, add nuance, and ensure it sounds like your organisation. Turns a 2-day writing task into 4 hours of editing.

What you get

Draft report sections with: (1) Narrative summary of programme and reach, (2) Data woven into storytelling (not just listed), (3) Beneficiary voices included via quotes, (4) Context about what the numbers mean, (5) Challenges and learning alongside successes. Always edited by humans before use - the AI generates the first draft, you make it accurate and authentic.

Before you start

• Impact data compiled: numbers, percentages, outcomes
• Programme context: what you do, who you serve, how it works
• Beneficiary quotes (anonymised if needed)
• Sense of what style you want (formal vs accessible, data-heavy vs story-led)
• Time to review and edit the output (never use unedited AI text)

When to use this

• Writing impact reports, annual reviews, or funder reports regularly
• You have the data but struggle to turn it into readable narrative
• Writer's block - you know what to say but can't get started
• Reports need to weave data into storytelling, not just list statistics
• You have beneficiary quotes and want to integrate them effectively

When not to use this

• Report is short and simple - quicker to write yourself
• Highly technical/specialist audience wants data tables not narrative
• You don't have time to review and edit properly (never send unedited AI text)
• Organisation policy prohibits AI-assisted writing
• Report contains sensitive information that shouldn't be shared with AI

Steps

1. 1

   Compile your data and context

   Gather everything: programme description, headline numbers (people reached, outcomes achieved, % improvements), breakdown by demographics if relevant, beneficiary quotes (anonymised), challenges faced, learning points. Put it all in one document to paste into the AI.

2. 2

   Define the style and structure you want

   Decide: length (500 words vs 2000 words?), tone (formal for trustees vs accessible for general public?), structure (chronological, theme-based, or outcome-focused?). The clearer your brief to the AI, the better the output.

3. 3

   Craft your prompt with data and context

   Create a prompt like: "Write a 1000-word impact report section for our youth mentoring programme. Include: [paste data]. Weave in these beneficiary quotes: [paste quotes]. Use an accessible, storytelling style that balances data with human stories. Structure: overview, who we reached, outcomes achieved, challenges and learning." Be specific.

4. 4

   Generate first draft

   Paste prompt into Claude or ChatGPT and generate. Read through the output. Does it capture your programme? Is the tone right? Are statistics used appropriately? Note what works and what doesn't - you'll refine in next step.

5. 5

   Iterate to improve the draft

   If output isn't right, ask for revisions: "Make this more data-focused", "Can you start with a beneficiary story instead?", "The tone is too formal, make it more accessible", "This section is too long, condense to 300 words". Iterate until you have a solid draft to edit.

6. 6

   Edit for accuracy and authenticity

   This is the crucial step: review every sentence. Check: are statistics accurate? Are quotes attributed correctly? Does this sound like your organisation? Are there nuances the AI missed? Add context the AI doesn't have. Fix any errors or misinterpretations. The human editing makes it usable.

7. 7

   Add elements AI can't generate

   Fill in gaps: specific examples the AI doesn't know about, recent developments, forward-looking statements about next steps, acknowledgments of partners/funders, organisational voice and personality. This is what makes it yours, not generic AI output.

8. 8

   Get colleague review before finalising

   Have someone else read it who knows the programme. Do they recognise it? Is anything inaccurate or misleading? Would beneficiaries recognise themselves in the quotes? Second pair of eyes catches what you miss. Never publish AI-assisted content without human review.

Tools

Resources