Advanced · Team and organisation · module 118 of 155
An AI use policy that people actually follow
Write a short, concrete policy with a traffic light for tasks and data, and keep it alive.
After this lesson
You create a use policy with concrete examples per colour, a reporting point and a fixed moment for review.
In view · Side by side
A traffic light for AI use
Allowed with the approved tool, such as rewriting your own text.
Allowed with conditions, such as no personal data and a colleague who proofreads.
Not allowed, such as pasting health data into an external tool.
Part 1 of 3 · Green
Your preference for less movement is on. Choose each part yourself.
Short and concrete beats long and vague
Hardly anyone reads a twenty-page policy. A usable AI policy fits on one or two pages and answers the questions staff really have: which tools may I use, which data may I enter, what must I check and whom can I ask (lesson 28)? Avoid general sentences such as use AI responsibly. Instead, write what that means for a specific task, with examples from your own work.
A traffic light for tasks and data
A traffic light makes choices quick to recognise. Green: allowed with the approved tool, such as rewriting your own text or summarising public information. Amber: allowed under conditions, such as a customer letter without personal data that a colleague proofreads. Red: not allowed, such as pasting health data into an external tool or letting AI decide about staff. For each colour, give two or three examples from your organisation. Treat doubtful cases as amber and ask about them.
A policy stays alive through reports and reviews
State in the policy where staff can ask questions and report errors or incidents, and make it clear that an honest report is not a problem. Appoint an owner who keeps the policy up to date (lesson 93). Plan a regular review, for example every six months, and also after a new tool, an incident or a change in the law (lessons 116 and 117). Explain why a rule exists; people follow rules better when they understand the reason. Test the policy with a few colleagues before you circulate it.
New practice example
From explanation to a usable result.
A good AI policy is short, concrete and uses examples. A traffic light helps staff choose quickly.
Fictional practice situationFictional: a municipal sports department is writing an AI policy. Staff write news items, answer questions from clubs and manage registrations of children for sports camps.
View a possible solution
This is how it can be done
Green: having a news item about a public event rewritten. Amber: a reply to a club without personal data, proofread by a colleague. Red: putting registrations with children’s names and medical information into an external tool. Contact point: the head of department. Review: every six months.
Why this works
Each colour gets examples from the team’s own work. That way an employee recognises their situation straight away.
Make it slightly more difficult
Find one task your team is unsure about, between amber and red. Which condition could move it to amber?
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
Each colour gets examples from the team’s own work. That way an employee recognises their situation straight away.
A step further
Find one task your team is unsure about, between amber and red. Which condition could move it to amber?
Try & discover · no obligation
Organise meeting notes
Categorise the notes as decision, action or open question. Make missing owners and deadlines visible.
Fictional AI policy of a customer service department. Green: allowed with the approved tool. Amber: allowed with conditions, such as no personal data and a colleague who proofreads. Red: not allowed.
- Green
- 0
- Amber
- 0
- Red
- 0
A traffic light only helps if each colour has concrete examples. If in doubt, treat the case as amber and check.
This trial works locally with fictitious data and does not count towards your lesson completion. It starts anew with a new lesson or reload.
A prompt to get you started
Help me write a one-page AI use policy for a fictional customer service team of ten staff members. Use a traffic light: green (allowed), amber (allowed under conditions) and red (not allowed). For each colour, give three concrete examples from customer service work. Add: approved tools (fill in: [tool]), what is always checked, where to report questions and errors, and when the policy will be reviewed.
Apply it
Your organisation has a thirty-page AI policy. Staff still ask every week whether they are allowed to have customer emails rewritten.
Which approach best fits this situation?
- Add another chapter to the policy.
- Make a short traffic light with concrete examples from the work, a reporting point and a review moment, and test it with colleagues.
- Ban all use of AI until everyone has read the policy.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
Which policy rule helps staff most with a concrete choice?
- Always use AI in a responsible way.
- Amber: a customer reply without personal data is allowed with the approved tool, if a colleague proofreads it before it is sent.
- AI is allowed, except when it is not.
Extra practice
Make a traffic light for your own team with three examples per colour. Show it to two colleagues and note which examples they would classify differently.
Green: rewriting your own draft email, summarising public product information, coming up with ideas for a newsletter. Amber: having a customer reply without personal data drafted that a colleague proofreads. Red: pasting a customer file with health information into an external tool, letting AI decide on a refund. Reporting point: the team leader. Review: every six months and after every incident.
