Skip to content

ChatGPT prompts: examples for everyday work

By the KI können editorial team7 min read

Written with AI and checked against the sources by a second AI. As of November 16, 2026. More in the notes on AI.

Good prompts are no secret knowledge. They tell the AI what to produce, for whom and within which limits. Here are examples for tasks that come up in almost every job. You copy them, put your details into the square brackets and check the result.

A woman stands at a standing desk typing on a laptop, a colleague with a notebook sits beside her, a board with notes on the wall.

Key points

  • The examples work for ChatGPT as well as for Claude, Gemini, Copilot or your company’s AI.
  • Replace everything in square brackets with your details. Leave out names and customer data.
  • Give the AI facts and limits, such as length, tone and what it must not make up.
  • Have the AI mark gaps or ask questions first instead of guessing.
  • Check every result before it goes out, and correct it one change at a time.

How to use the examples

A prompt is your instruction to the AI. The more precise it is, the less the AI has to guess. Every example below follows the same pattern: what to produce, for whom, what for, what goes in and which number applies, such as the length. Why these five parts in particular is explained in the guide How to write prompts.

  1. Pick an example that fits your task.
  2. Replace everything in square brackets with your details.
  3. Delete what doesn’t fit and add what is different for you.
  4. Check the result before you use it.

In the trial lesson you put your first instruction together and get feedback. No account, in a few minutes. Start the trial lesson

Writing and shortening e-mails

Aylin works in customer service at a furniture store. Deliveries are often delayed, and every notice costs her time. She gives the AI the facts and lets it write only the form.

Sample instruction
Write an e-mail to a customer whose delivery is delayed.
Facts: She ordered a dining table, planned for [date]. New date: [date], time window [time] to [time]. Reason: the manufacturer is delivering late.
Tone: friendly, factual, formal. 120 words at most.
Apologize once, not several times. Offer that she can suggest another date.
Don’t invent vouchers, discounts or promises that are not stated here.

The last sentence matters. Without it, an AI likes to offer a discount that doesn’t exist. For long e-mails you wrote yourself, the opposite helps: shortening.

Sample instruction
Shorten my e-mail below to 80 words at most.
Keep all dates, figures and requests. Don’t change any facts.
Mark anything unclear to you with [?] instead of guessing.
[Your e-mail without names]

Explaining and rewriting texts

Many texts already exist; they are just too long, too technical or in the wrong tone. Mateo is the building superintendent of a residential complex and has to make the house rules understandable for the notice board. The house rules are not confidential; they hang in the stairwell anyway.

Sample instruction
Rewrite the following section of the house rules in plain language.
Audience: all tenants, including people who are still learning the language.
Short sentences, 12 words per sentence at most. One rule per line.
Don’t change any times or rules. Don’t leave anything out.
At the end, list every place where you were unsure.
[Section of the house rules]

Then he puts the original and the new version side by side and checks every rule. When simplifying, a fact easily slips, such as “from 10 p.m.” turning into “at night”.

Changing the tone

Sample instruction
Here is my reply to a complaint. It sounds too irritated.
Make it calmer and friendlier without changing the content.
Give me two versions: a short one with three sentences at most and a longer one.
[Your reply without names]

Minutes and summaries

Miriam is a clerk at a city administration and writes the minutes after every team meeting. She notes bullet points with initials instead of names and has the AI turn them into a text.

Sample instruction
Turn my bullet points into minutes of the team meeting on [date].
Structure: participants (initials only), topics, decisions, tasks with initials and deadline.
One page at most. Include only what is in the bullet points.
Where a deadline or a responsible person is missing, write [missing] instead of making something up.
[Your bullet points]

The [missing] mark shows Miriam at once what she still has to clarify. She replaces the initials with names only in the finished document, outside the chat.

Summarizing long texts

For public texts such as a brochure, an overview of a standard or a trade article, this instruction works. Upload internal documents only to an AI your company has approved for them.

Sample instruction
Summarize the following text in five points at most.
Audience: colleagues with no prior knowledge. One sentence per point.
Then name three questions the text leaves open.
For each point, quote the passage you are relying on.
[Text]

Checklists and plans

Rafael is a shift supervisor at a print shop. New colleagues shouldn’t miss anything important in their first week. He has the AI draft a checklist and adds his knowledge of the business.

Sample instruction
Draft a checklist for the first working week of new employees in the production area of a print shop.
Organize it by day, Monday to Friday. Six points per day at most.
For each point: what to do and who shows it (placeholder [person]).
Occupational safety and safety instruction belong on the first day.
Don’t include any details about laws or deadlines; I will check those myself.

What exactly the safety instruction has to cover is decided by the business, not by the AI. The list is a framework that Rafael goes through with the occupational safety specialist.

Organizing a weekly plan

Sample instruction
Here are my tasks for next week with estimated duration: [list].
Fixed appointments: [appointments]. I work [hours] hours a day.
Spread the tasks over the days. Plan half an hour of buffer per day.
Tell me if the time isn’t enough, and suggest what I could postpone.

Setting up a spreadsheet

Sample instruction
I want to keep track in Excel of which devices in our workshop need to be inspected and when.
Suggest the columns, eight at most, each with a sample value.
Tell me which column I should format as a date and how to have due inspections highlighted in color.

For formulas and analyses, the AI needs more details, such as the program, the language and the columns. How that works is in the guide AI and Excel.

Let the AI ask first

Sometimes you don’t know exactly what you need yet. Then an instruction helps in which the AI asks first instead of starting to write right away.

Sample instruction
I want to write a notice for our break room so that dishes stop being left out.
First ask me five questions at most that you need for a good notice.
Write the notice only after I have answered.

In the same way, you can ask the AI to push back. That is useful before you present an idea to your team.

Sample instruction
Here is my proposal for new rules on vacation planning in our team: [proposal].
Name the three strongest objections my colleagues might have.
For each objection, suggest an answer or a change.

Improving: one change per step

The first result rarely fits completely. Instead of rewriting the whole instruction, say exactly what should change: where, what it says now and what it should say.

  • “The second paragraph says ‘soon’. Write the date [date] instead.”
  • “The list has nine points. Shorten it to six and keep the first three unchanged.”
  • “The ending sounds too much like an ad. Replace it with a factual sentence with our phone number.”

Change only one thing per message. Otherwise you won’t know which change did what. If it gets worse, go back to the last good version and try another way.

Check before you use it

Whichever example you use, the result is a draft. Before it goes to customers, your team or your boss, go through these questions:

  1. Do all figures, dates and names match your details?
  2. Has the AI added something you didn’t say, such as a promise?
  3. Does the tone fit you and the person reading it?
  4. Is the requested length kept?
  5. Is there anything confidential in it?

Whatever is wrong, correct it yourself or with a targeted follow-up instruction. Which data may go into an AI chat at all is explained in the guide Data protection and AI tools.

How to practice this

Examples help you get started. You become confident when you write instructions for your own tasks and check the results carefully. In the learning app “KI können” you practice this in short lessons with examples from your job and put the instructions to work in your own AI. The trial lesson needs no account.