What You Will End Up With
You will have a reviewed draft report that a program officer can read quickly and trust. It shows that the money went where the application said it would, sets budget against actuals, and tells the story of the period with setbacks included. The AI does the first pass of drafting. You supply the evidence and make the calls. Scoring the draft against a rubric before it leaves your desk helps catch weak spots before they reach the funder.
Before You Start
An assistant cannot invent your results or check your spending, so collect the evidence first.
- The original application and the signed agreement. These define what was promised, and financial accountability is measured against them.
- A budget-versus-actual summary. Keep year-to-date actuals current so the numbers are ready whenever a deadline lands.
- Indicator updates from the approved results framework. Real figures, with targets beside them.
- Notes from program staff. Context, planned versus actual results, outcomes, risk and procurement issues, coordination with other actors, cross-cutting issues such as gender or environment, and lessons learned.
Then settle two questions. Who is the primary reader, and what do they need the report for? A compliance-minded reader wants different framing from a general one. And what does this funder actually require? Some want no written report, some accept a verbal update, and others insist on their own template and timeline.
Steps
-
Draft one core report. Most funders ask for the same underlying information in different words, so write one strong core report and adapt it later. Give the assistant your indicator data, staff notes and results framework. Ask it to draft sections on program and context status, planned versus actual results, outcomes, management issues, coordination, and lessons. By hand, check that the account is balanced. Setbacks belong next to successes. Also check that each section traces activities to outputs and outputs to outcome-level change.
-
Tie the money to the promise. Give the assistant the budget-versus-actual summary and the application. Ask for a financial section showing that funds were used for the purpose stated. By hand, compare every figure with your finance records, and make sure the narrative agrees with the numbers. Many funders that supply a financial template expect exactly that consistency. If a variance is significant, add a short exception note explaining it.
-
Reorder for the reader. A program officer with many grants tends to read in a fixed order, not top to bottom. Ask the assistant to lead with the most important findings and to open each section with the result, then the evidence. By hand, check that the report also shows how your work fits into the wider country or sector effort.
-
Adapt to the funder's template. Paste in the template and your core report (after removing anything confidential, see Using AI Safely), and ask the assistant to excerpt and reshape sections to fit each question. By hand, look for anything it skipped. Funder templates commonly ask for an update on the relevant indicators in the approved results framework, not a narrative summary alone. For a restricted project grant, expect to pull more detail from program staff than a general operating support report would need.
-
Tighten and add visuals. Ask the assistant which sections would work better as a chart. One well-chosen chart can replace a page of text, and complex visuals belong in a labeled annex. Ask it to cut filler, not evidence. By hand, confirm no claim is stated more strongly than the data supports.
-
Score it against the rubric. Ask the assistant for a preliminary score on each dimension below, using a numbered scale you set in advance with 1 as the lowest score, then have a person make the final call.
| Dimension | What to check |
|---|---|
| Indicator results | Real figures against targets, gaps explained, cumulative progress shown |
| Data evidence | Each result tied to a named data source, data quality limits disclosed |
| Narrative logic | Explains what happened and why it matters, connects activities to results |
| Honest disclosure | Challenges stated with root causes and specific mitigation actions |
| Applied learning | Specific lessons, and how they will change the program's next steps |
For any dimension scored 1 or 2, add a brief explanation of the problem and a concrete example of how to revise it. Where you can, have a peer, an advisory panel or a reference group look at the draft before it goes out, and do it while the draft is still taking shape, not only at the end.
Using AI Safely
Treat everything the assistant returns as a draft, and check it by hand. A report must never state a claim more strongly than the data supports.
- Keep confidential information out of public AI tools. Remove or anonymize anything confidential before pasting.
- Check every figure against its source. Each number, percentage and date goes back to the budget-versus-actual summary or the results framework.
- Compare the story with the agreement. Read the narrative against the application and signed agreement to be sure the assistant has not invented a goal or activity that was never promised.
If the assistant drops a setback to make the period sound better, put it back. A credible report gives a balanced account, with setbacks alongside successes.
Common Mistakes
Listing activities. A report that only says what was done tells the reader nothing about why it mattered. Explain what happened and why it matters, linking activities to outputs and outcomes.
Reporting only successes. A credible report includes what went wrong. Name the challenge, its root cause and what you are doing about it.
Describing indicators in prose. Give the figure against the target, explain any gap, and show cumulative progress.
Filing the report and forgetting it. Share the final version with the field staff and partners who contributed. Then use it as the starting point for the next work plan, carrying unresolved challenges forward.
Overstating. If the data cannot carry a claim, soften the claim or find the data.