How to use these prompts
Whoever reads a report at work wants the conclusion right away, not at the end. A useful document puts up front what happened and what to do, and pushes the details further down. The prompts below follow this logic: first the decision summary, then the body. That way someone with thirty seconds gets it, and someone with ten minutes can dig deeper.
One rule that holds for the entire work cluster: AI provides structure and prose, you provide the numbers and the assessments. If you make it invent a revenue figure or a percentage, it will invent one. Give it the real data and ask it to organize it, never to produce it. And check every figure before the document leaves your computer.
The prompt library
Build the report's scaffolding
Help me structure a report on this topic before I write it.
Give me a reasoned outline: which sections are needed, in what order, what
goes in each one, and where to put the decision summary up front. Think about
who reads it and how much time they have. Don't write the content yet.
Topic of the report, audience, and the decision it has to support:
Write the summary for those in a hurry
Write the opening summary of this report for an executive who has
thirty seconds: what happened, what it means, what I propose to do.
Five lines maximum, at the top of the document. No preamble, get to the
point. Use only the data I give you, don't add any.
Content of the report:
From raw data to a readable report
Turn these data and notes into a professional report. Structure:
summary up front, then the results with the numbers I give you, then the
implications, finally what I propose. Don't invent figures: if a piece of
data is missing, write in square brackets what is needed. Dry and
objective tone.
Data and notes:
Adapt the same report to two audiences
I have this report. Give me two versions: a technical one for those who know
the subject, a simple one for those who decide but aren't in the field. Same
facts, different depth. In the simple version, explain the technical
terms the first time they appear.
Report:
Final review before delivering
Review this document the way someone who has to approve it would. Tell me:
where it isn't clear what I'm proposing, where a statement lacks the data that
supports it, where the tone is too long-winded. Flag the vague phrases like
"the results are positive" that should be quantified. Don't
rewrite everything: point me to the weak spots.
Document:
A concrete example
Elisa has to hand the director a quarterly report on how her department is doing. She has the numbers scattered across three spreadsheets and two weeks of notes. She starts from the scaffolding: the AI suggests she put a summary up front with the key result and her proposal, then the data by area, then the risks.
She fills in the "from raw data" prompt, pasting in the real figures and the notes. The AI gives her back an orderly report and flags with square brackets two spots where a year-over-year comparison figure is missing. Elisa retrieves those two numbers and inserts them. Then she runs the draft through the final review: the AI points out that "sales improved" doesn't say by how much, and she turns it into "sales grew by nine percent". The director reads the first five lines and already has the picture. Elisa wrote the report; the AI removed the two hours of formatting and acted as her first critical reader.
When it does NOT work (and how to fix it)
If it slips in data or conclusions you didn't provide
Empty space pushes the AI to fill it. Block it in the prompt: "use only the data I provide, don't add figures, percentages or sources; if a piece of data is missing, flag it in square brackets". That way it marks the gaps instead of inventing them, and you know where to step in.
If the report is bloated and repetitive
The AI tends to pad to seem thorough. Impose the cut: "two pages maximum, every sentence must add a piece of information, eliminate repetitions and roundabout phrasing". A report that reads quickly gets read; a long-winded one gets badly skimmed by whoever receives it.
If the tone is too enthusiastic for a serious document
Drafts often come out with brochure-like enthusiasm. Correct it: "sober and objective tone, no promotional adjectives, report the facts and let them speak". A report convinces with numbers, not with adverbs.
A tip from someone who actually uses it
Write the opening summary yourself, even badly, before asking the AI for help: it forces you to get clear on what you're really saying. Then hand it to the AI to polish. If instead you start from "write me the report", you delegate the thinking too, and it shows: out comes a correct document that says nothing. The useful effort is deciding what matters; you do that, the tool accelerates the rest.
Frequently asked questions
Can I have the AI write an official report that carries my signature?
The draft yes, the responsibility stays yours. You answer for what you sign, so every figure must be verified and every conclusion must be yours, not the one the AI guessed. Use it for structure and speed; the final validation, line by line on the numbers, can't be delegated.
How do I keep the report from looking "written by AI"?
Three moves: put a concrete piece of data that only you know in every section, cut the enthusiastic adjectives, and rewrite the opening in your own words. The AI's recognizable stamp lies in the polished, generic sentences; just anchor them to specific facts from your case and they disappear.
Does a longer, more detailed report make more of an impression?
It's the opposite, for someone who has to decide. A long document often signals that whoever wrote it didn't know what to cut. The ability to summarize is what separates a professional report from a volume of notes: the five-line summary up front is worth more than twenty pages nobody finishes. Brevity done well is harder and more effective.