AI & automation · Security · SMEs
AI rules at work: what your team can put into ChatGPT and co.
Your team is probably using AI already; the only question is with which data. A simple traffic light and a one-page AI rule bring clarity without slowing work down.
Author: Christoph SteinerManaging director & founder
Published on 4 min read
Created with AI assistance, reviewed by Christoph SteinerHow we use AI
Your team can use AI tools such as ChatGPT, Copilot or Gemini, as long as there is a clear rule about which data goes in. General texts, ideas and public information are fine. Internal documents and customer data only belong in tools your business has checked and approved. Passwords, bank details and personnel files stay out altogether. Ideally you put this on a single page, so everyone on the team knows where they stand.
Why this shouldn't wait
In many small businesses, AI is already part of the daily routine, even though nobody ever decided on it. The apprentice rewrites an email to a supplier, a colleague in sales has a long quote summarised, accounting asks for a spreadsheet formula. Often this happens through private accounts on private phones.
That's no reason to panic, but it is a reason to act. Without a rule, each person decides for themselves what they paste into a text box. And anyone who doesn't know what's allowed either makes mistakes or avoids AI altogether and misses out on the time it could save.
A clear rule solves both: it protects your data and gives your team the confidence to use AI where it helps.
The traffic light: what goes in and what doesn't?
Instead of a long policy, a simple traffic light is enough for most businesses.
Green – fine to use:
- General texts without names, such as drafts for job ads, social media posts or product descriptions
- Public information that is already on your website
- Questions about formulas, wording, translations or ideas for the next team meeting
Amber – only in the approved company tool:
- Internal documents such as quotes, meeting notes or price lists
- Texts that mention customers or suppliers by name
Red – stays out:
- Passwords, login details and bank details
- Health data, payslips and personnel files
- Trade secrets and confidential contracts
A simple rule of thumb for the team: anything you wouldn't want to see printed out in your shop window is never green.
Private account or company account: the subtle difference
Many AI providers have plans for individuals and separate offers for businesses. They often differ exactly where it matters to you: whether your input may be used for training, who in the business manages the accounts, how long data is stored and where it is processed.
Read the terms of your tool, or have someone explain them, before you approve it for the amber level. Once personal data is involved, such as customer names, you need a tool where the provider only processes that data on your behalf, backed by a data processing agreement (Art. 28 GDPR). Business plans usually offer one; private accounts don't.
In practice: one tool for everyone that you know and manage is better than five private accounts you know nothing about. When someone leaves the business, you lock a company account centrally. A private account takes the chat history with it.
Check the results: AI sounds sure even when it's wrong
The second half of any AI rule isn't about what goes in, but what comes out. AI tools write fluently and convincingly, even when a detail is simply wrong. A made-up figure, a deadline that doesn't exist, a quote nobody ever said: it happens.
So the rule is:
- Anything that goes to customers is read by a person first.
- Figures, prices, legal and technical points are always checked against the source.
- Whoever sends or publishes an AI text remains responsible for it, not the tool.
That way, AI stays what it should be: a quick first draft, not an unchecked final result.
One page is enough: how to write your AI rule
You don't need a legal department to get started. One page with these six points covers most of it:
- Approved tools: Which AI tools do we use, and with which account?
- The traffic light: What is green, amber and red, with examples from your daily work?
- Check before sending: Who reviews results before they go out?
- Transparency: Where do we say that AI was involved, for example in images or automatic replies on the website?
- Contact person: Who do you ask when you're unsure?
- Review: When do we look at the rule again, for example every six months?
Go through the page briefly with your team and ask which tools are already in use. That usually tells you more than any survey. It also shows your team that AI is welcome, as long as everyone knows where the limits are.
Important: this rule is no substitute for legal advice. If you work with particularly sensitive data, for example in healthcare, get expert advice as well.
The next step: AI where your data belongs
The traffic light quickly shows where the real value lies: the most interesting tasks are often amber. Preparing quotes, sorting customer enquiries, searching internal documents. That calls for solutions where it is clear where the data goes and who can access it.
At Proplex, we build AI automation for businesses: built into your workflows and planned so your data stays where you want it. You can see what else we do under services, and find more articles on AI and digital topics on our blog.
Start small: write down your traffic light this week, talk it through with your team, and you'll quickly see which amber tasks are worth automating safely.


