Deep dive · module 54 of 100
Team agreements on prompts
Make agreements so that the team works consistently and safely with AI.
After this lesson
You can formulate three concrete agreements regarding prompts and control.
In view · How it connects
Three agreements for every team
Record which applications you support.
Determine what may and may not be included in prompts.
Assign a responsible person before the output is used.
Part 1 of 3 · Which tasks?
Your preference for less movement is on. Choose each part yourself.
Which tasks are okay?
Agree on which tasks AI is allowed to support.
What is not allowed in the prompt?
For example: real customer data, passwords, internal figures.
Who checks?
Record who reviews the result before it is used.
New practice example
From explanation to a usable result.
Team agreements make it visible which tasks, data, and checks are appropriate for AI use.
Fictional practice situationA fictitious association wants to use AI for public activity announcements and for internal complaints. The complaints contain personal details.
View a possible solution
This is how it can be done
Agree: public announcements may use confirmed source facts; practice complaints are fictitious. For real complaints, an assessment of data and permitted tools must come first. A designated colleague checks publications.
Why this works
The same tool may require different agreements for different tasks. The document should help someone decide what the next safe step is.
Make it slightly more difficult
Write a single rule for a doubtful case: whom do you consult, what do you omit in the meantime, and where do you record the decision?
Extra practice, at your own pace. You do not need to submit anything. This does not count towards an automatic assessment.
Your workplace · learning by doing
Make something. See what changes.
Checking the AI connection… You can already view the written example.
Present your assignment to Ende.
Read the answer, change one thing, and compare.
Previously held
New version
You remain in control
What do you want to improve in your next attempt?
On this device only. AI answers remain only in this page memory. This is self-review, not a grade.
Why this approach works
The same tool may require different agreements for different tasks. The document should help someone decide what the next safe step is.
A step further
Write a single rule for a doubtful case: whom do you consult, what do you omit in the meantime, and where do you record the decision?
A prompt to get you started
Help me formulate three short team agreements regarding the use of AI prompts. Topics: permitted tasks, forbidden input, and control.
Apply it
Two colleagues use the same prompt differently. One sends draft emails immediately, the other checks the facts first.
Which approach best fits this situation?
- Establish shared agreements on checking and usage.
- Let everyone decide for themselves whether a check is necessary.
- Share only the prompt text without agreements.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
Why are team agreements about prompts important?
- To work consistently and safely.
- Because AI otherwise does not work.
- Only for large companies.
Extra practice
Write three concrete agreements that your team could use.
1. No real client names in prompts. 2. Always check for fabricated promises. 3. Always have important emails read by a colleague.
