Before you use AI at work
Know what you may paste, which account to use and whom to ask, so work data stays where your company allows.
About 8 minutes (editor's estimate) · Updated 2026-10-16
What you will learn
- The three kinds of work information: fine to paste, clean first, never paste
- Why a company-approved tool and a personal account are not the same
- What privacy settings control
- What to ask IT or your manager before you start, and why you stay responsible for what is sent
The idea
What you type or upload into an AI tool is usually sent to the tool's provider. At work, that material often belongs to your company, its clients or your colleagues. So before the first prompt, sort it into three kinds.
- Fine to paste: public information, such as a published web page, and text with nothing about clients, colleagues or the business.
- Clean first: work material that is fine once you remove or replace names, contact details, order numbers, prices and anything else that points to a person or a client. Most meeting notes and customer emails are here.
- Never paste: passwords, bank and ID numbers, health, salary or HR details about anyone, and anything your company's rules keep out of AI tools. If you are unsure, treat it as never paste until you have asked.
Next, the account. Many AI products offer both a personal account and a work account set up by your company, and what happens to your data depends on which one you use. A work account usually comes with terms your company agreed with the provider. Its administrators can control many settings and may be able to keep and review what you ask. A personal account is between you and the provider, and your company usually has no control over what is in it. So company material goes only into the approved tool and account.
Many tools also have privacy settings that control whether your chats may be used to improve the provider's models and whether your chat history is kept. On a work account, your company may have set them for you. Look at them before you start.
Last, the tool can draft, but you are the sender. Read and check everything before it goes out.
Watch
Explains five ways using AI can get you into trouble at work (unapproved tools, leaking sensitive data, passing on made-up answers, prompt injection and unapproved AI agents) and how to avoid each one.
Example
Lena works in the office of a delivery company. She wants help with a calm reply to a customer whose parcel arrived late. Her phone has a chatbot on a personal account. First she sorts what the reply needs:
Fine to paste: our delivery terms, published on our website
Clean first: the complaint (customer name, address, order number)
Never paste: the screenshot of the customer's bank transfer
She asks the office manager. The answer: use the approved assistant with her work account, not the chatbot on her phone. Customer names and order numbers are allowed in it, bank details never. Unclear cases go to the data protection contact.
Lena signs in and pastes the complaint without the screenshot. The draft promises a refund that only her manager can approve, so she removes that sentence and sends the reply herself.
Practice now
- Pick one work task, such as a reply or a summary.
- List the information it needs and sort each item: fine to paste, clean first or never paste.
- Find out who decides about AI tools where you work: IT, your manager or a data protection contact.
- Send them the message below today. Describe your material; do not attach it.
- Until they answer, use only what is fine to paste. When they do, note the tool, the account, what is off limits and who to ask, then look at that account's privacy settings.
At the end you should have a sorted list, a question with the right person and, once they reply, your company's rules in one place.
Copy this template
Subject: Using AI tools for my work
Hi [name],
I would like to use an AI tool to help with [task, for example replies
to customer emails]. Before I start, could you tell me:
1. Which AI tools are approved for work?
2. Which account should I use, and how do I get access?
3. What information must never go into an AI tool?
4. Which settings should I check, such as chat history or whether
chats may be used to train the provider's models?
5. Who should I ask when I am unsure about a specific case?
The material would be [describe it without pasting it, for example
"customer emails with names and order numbers"].
Thank you,
[your name]If the answer is "it depends", ask for one example of what is allowed and one of what is not.
Check the output
- Do you know the approved tool, the account to use and who to ask next time?
- Do you know what is off limits, in plain words and not only "be careful"?
- Before you paste, is every item fine or cleaned, with nothing from the never paste list?
- Are you signed in with the work account, and did you read the AI's text before anyone else saw it?
Common mistakes
- Using the account that is already open. The chatbot on your phone is often a personal account, outside your company's control.
- Thinking a paid plan makes it a work tool. A personal plan you pay for is still a personal account.
- Cleaning only the names. Order numbers, email signatures and file names can still point to a person or a client.
- Letting the tool speak for you. A draft can promise a refund or a date nobody agreed to. You send it, so you check it.
Finished the practice?
Progress is saved in this browser only. Clearing your browser data removes it.
Sources
- Data, Privacy, and Security for Microsoft Copilot (Microsoft, read 2026-10-16)
- Generative AI in Google Workspace Privacy Hub (Google, read 2026-10-16)
- Google: 13594961 (Google, read 2026-10-11)
- Data controls in ChatGPT (OpenAI, read 2026-10-10)