Before you open any AI tool
Welcome back! In Level 1 we looked at what AI can and can't do in customer service. Now for a question that comes before you type a single prompt: which tools are you actually allowed to use at work?
It sounds like admin. It isn't. Get this right and you can use AI with confidence. Get it wrong and you could put customer information somewhere your organisation can't see or control, without meaning to.
And you wouldn't be alone in getting it wrong. In Microsoft and LinkedIn's 2024 Work Trend Index (May 2024, a survey of 31,000 people in 31 countries), 78% of people using AI at work said they brought their own AI tools rather than ones their employer provided. Most of them meant well. Even so, that's exactly the risk we want you to avoid.
Same name, different tool
Here's the part that catches people out. ChatGPT, Gemini, Claude and Copilot all come in two very different forms.
There's the personal (consumer) account. That's the one you sign up for at home with your own email address. Then there's the workplace (enterprise or business) version. Your organisation buys that one, sets it up and manages it.
Think of it like a company car and your own car. Both get you from A to B. But the company car is insured for work, maintained by the business and covered by its rules. If you use your own car for a work delivery, you're on your own if something goes wrong.
So what's actually different? Let's look at what the vendors themselves say.
With Microsoft 365 Copilot, used through a work account, Microsoft says your organisation's prompts and responses aren't used to train its foundation models, and the data stays inside your organisation's Microsoft 365 setup. ChatGPT Business and Enterprise: OpenAI says it doesn't train its models on business data by default. Gemini in Google Workspace: Google says Workspace data isn't used to train its models outside your organisation without permission. Claude for Work (Team and Enterprise plans): Anthropic says commercial data isn't used to train its models.
Now compare that with personal accounts. On consumer ChatGPT plans, your conversations can be used to improve OpenAI's models unless you switch off a setting called "Improve the model for everyone". Since Anthropic changed its consumer terms in 2025, people on Claude's Free, Pro and Max plans choose whether their chats are used for training.
"But I switched training off!"
Good question, and a really common one. Say you've turned off training on your personal account. Does that make it fine for work?
Sadly, no. Training is only one part of it. With a personal account your employer has no contract with the vendor, can't manage your settings, can't check what was shared, and can't delete it if a customer asks. The data sits in your account, not the company's.
An approved tool is approved because someone at your organisation has checked it, signed an agreement for it and set it up properly. Fiddling with your own settings doesn't give you any of that.
One more catch. The logos look identical. So how do you know which version you're in? Check that you're signed in with your work account, through the link or app your IT team gave you. If you aren't sure, ask.
Finding and reading your policy
So where are the rules written down? Most organisations have an AI policy, or AI rules inside a wider acceptable-use policy. Try your intranet, staff handbook or IT help pages, or just ask your team leader where it lives.
When you find it, you don't need to memorise every word. Read it looking for a handful of answers:
- Which AI tools are approved, and which are banned?
- What kinds of information can and can't go into them?
- Which tasks is AI allowed for, and which are off-limits?
- Does a human have to check AI output before a customer sees it?
- Do you need to tell anyone when you've used AI?
- Who do you ask when you're unsure?
Try summarising your policy in five or six lines of your own words. If you can't, that tells you where to dig a bit deeper.
When the policy doesn't cover your task
Policies can't predict everything. Say you want to use AI to tidy up notes from a tricky call, and the policy just doesn't mention it. What do you do?
First, pause. Don't fill the gap with a guess. Then go to your team leader with specific questions, not just "Can I use AI?"
You might ask: Which approved tool should I use for this? What information can I include, and what must I leave out? Does someone need to review the output before it goes anywhere? Do I need to record or mention that AI helped? Is there a template or example I can follow?
Then write down the answer you get, so you and your teammates have it next time. Specific questions get you useful answers. They also show your team leader you're taking this seriously, and that does you no harm.
Your turn
This week, find your organisation's AI or acceptable-use policy. Write your own five-line summary. Note which AI tools are approved and how you sign in to them. Then write down one task the policy doesn't clearly cover, plus the questions you'd ask about it.
Key takeaways
- Personal and workplace versions of the same AI tool are different products with different protections.
- Workplace versions of Microsoft 365 Copilot, ChatGPT, Gemini and Claude don't use your organisation's data for training by default. Consumer accounts work differently.
- Turning off training on a personal account doesn't make it an approved work tool.
- Know where your policy is, and be able to summarise its key rules.
- When the policy is silent, pause, ask your team leader specific questions and write down the answer.