Use this guide before you start using AI chat tools for work tasks, or if you already use them and want to do it more safely. Have your employer's IT or data policy to hand if one exists.
Step by step
Check the rules first
Find out whether your employer allows AI tools and which ones are approved. Some organisations provide a business version with stronger data protections. If there is no policy, ask your manager or IT contact before using one for work.
Understand where your data goes
Read the tool's settings and privacy information to see whether conversations are stored or used to train the service. Switch off history or training options where that is allowed and appropriate. Assume anything you type could be seen by the provider.
Keep confidential details out
Do not paste personal data about customers or colleagues, passwords, financial details or anything marked confidential into a tool that is not approved for it. Replace names and figures with placeholders when asking for help with structure or wording. UK data protection law still applies to what you share.
Give clear, specific instructions
Say what you want, who it is for, the tone and the length. Provide the relevant background in your own words. Clear requests produce more useful drafts and fewer errors.
Check everything before using it
AI tools can state wrong facts confidently, invent sources and miss context. Verify figures, names, dates and any legal or technical claims against a reliable source. You remain responsible for anything you send or publish.
Be open about how you used it
Follow your organisation's rules on telling people when AI helped produce work. Keep your own judgement in decisions that affect people, such as hiring or complaints. Record what you used it for if your team needs an audit trail.
Ready-to-use checklist
- Employer policy checked
- Approved tool identified
- Privacy and history settings reviewed
- No personal or confidential data pasted
- Facts and figures verified
- Output edited into your own words
- Use disclosed where required
Practical tips
- Ask the assistant to list its assumptions so you can spot where it may be wrong.
- Use AI for first drafts, summaries of your own notes and checklists rather than final answers.
- Keep a short list of prompts that work well for your regular tasks.
Common problems
I pasted something confidential by mistake.
Delete the conversation if the tool allows it and tell your manager or data protection contact straight away. They can judge whether it needs to be reported. Acting quickly matters more than getting it perfect.
The assistant gave me a source that does not exist.
This is a known weakness. Never cite a source you have not opened and read yourself, and search for the original document directly.
My colleague uses an unapproved AI tool for client work.
Raise it calmly and point to the policy, or mention it to your manager if the risk is serious. The aim is to protect clients and the business, not to blame anyone.