Before You Open Any AI Tool
Welcome back! In Level 1 we looked at what AI can and can't do for you at work. Now there's a question you need to answer before you type a single prompt: which tool are you allowed to use?
It sounds like a small thing, doesn't it? It isn't. Two tools can look the same on screen and still handle your data in very different ways. We'll work through it together.
Think of It Like the Office Keys
Here's a picture to keep in mind. Your office gives you a set of keys. You can get into the building, the stationery cupboard and maybe your manager's office. You don't get a copy cut at the shop round the corner and hand it to a stranger.
AI tools work the same way. Your organisation has probably "issued" some tools and not others. The approved ones come with agreements about how your data is protected. A tool you sign up for yourself has no such agreement with your employer, so it's a bit like that copied key.
Step One: Find Your Policy
So where do you start? Look for your organisation's AI policy. It might be its own document, or it might sit inside the IT policy, the acceptable use policy or the data protection policy. Try the staff intranet, your employee handbook or the IT help desk.
What if you can't find one? Ask. A short message to IT or your line manager is enough: "Do we have guidance on using AI tools like ChatGPT or Copilot? I'd like to make sure I'm using them properly." Nobody will think less of you for asking. It shows good judgement.
When you do find the policy, don't try to memorise all of it. Pull out the rules that touch your day: which tools are approved, what kinds of information you mustn't enter, whether you need to say when you've used AI, and who to contact with questions.
Consumer vs Work Accounts: Why It Matters
This is the important part. Most AI providers offer two kinds of plan. One is a consumer plan that you sign up for yourself, free or paid. The other is a business or enterprise plan that your organisation buys and manages.
The big difference is whether your conversations can be used to train future versions of the AI. Let's look at two examples, based on the providers' own pages as of October 2026.
ChatGPT (OpenAI). OpenAI's help centre says that for individual services like ChatGPT, it may use your content to train its models unless you opt out. You can switch this off in settings, and Temporary Chats aren't used for training. On ChatGPT Business and Enterprise, OpenAI says it doesn't train on your business data by default.
Claude (Anthropic). On the consumer plans, Free, Pro and Max, you choose whether your chats are used to improve Claude. If you leave that setting on, Anthropic says it may keep de-identified data for up to five years in its training pipelines. Incognito chats aren't used for training. The work products, such as the Team and Enterprise plans, fall under Anthropic's commercial terms, which say it may not train models on customer content.
The same pattern shows up elsewhere. Microsoft says prompts and responses in Microsoft 365 Copilot Chat signed in with a work account are covered by enterprise data protection and aren't used to train its foundation models. Google says content in Gemini for Workspace isn't reviewed by humans or used for model training outside your domain without permission.
Do you see the pattern? The work version comes with a contract between your employer and the provider. The consumer version is an agreement between the provider and you as a private person.
One more thing: these pages change. That's why you should check the provider's current help page yourself rather than relying on what a colleague remembers.
Is This Tool Approved for This Task?
Let's make it practical. Say your manager asks you to tidy up the minutes from a board meeting. You've got ChatGPT on your personal phone. Should you use it?
Run through three quick questions. Is the tool on your organisation's approved list? Are you signed in with your work account, not a personal one? And does the policy allow this kind of information in that tool?
If all three are a clear yes, go ahead. If any of them is "I'm not sure", stop and ask. Usually that means your line manager, IT or your data protection lead.
And notice something: a tool can be approved for some jobs and not others. Your policy might allow Copilot for drafting a general letter but not for anything involving HR files. Approval depends on the tool and the task.
Your Turn
This week, find your organisation's AI or IT policy, or send that short message asking for it. Write down which tools are approved and which account you should sign in with. Then pick one task you do every week and decide whether the policy covers it.
Key Takeaways
- Find your organisation's AI policy, or ask IT or your manager for it, and pull out the rules that affect your daily tasks.
- Consumer and work accounts can handle your data very differently, especially when it comes to model training.
- OpenAI and Anthropic both say they don't train on business plan data by default, while their consumer plans may use your chats unless you change the setting.
- Check the provider's current help pages yourself, because policies change.
- Before you use a tool, ask: is it approved, am I on my work account, and is this task allowed? If you're unsure, ask first.