Guardrails for administrative assistants

When you finish, you can decide what never goes into an AI tool, how to check output, and when to say no.

You see what others cannot

Assistants often see salary talks, medical leave, legal letters, board drafts, passwords and plans that are not public yet. That access is built on trust. A consumer AI tool is a place where text may be stored, reviewed or used in ways you do not control, depending on the product and your settings.

So the first rule is simple. Do not paste anything into an AI tool that you would not be comfortable seeing in the wrong inbox. Unless your employer has approved a specific tool for a specific kind of information, treat the tool as public.

What never to paste, and how to work around it

Never paste passwords, bank or card numbers, ID numbers, health details about a colleague, performance or disciplinary notes, contracts not yet signed, or anything marked confidential. Also be careful with a client’s name next to their problem. Even without a name, a few details together can point to one person.

You can still get help by removing details. Instead of “Maria Lopes in finance is on sick leave until the 14th”, write “a colleague will be out for two weeks”. Use placeholders like [Client A]. If you cannot do the task without the private details, do it by hand or ask whether your employer has an approved tool.

Accuracy, bias and tone

AI tools can sound sure while being wrong. They can get a time zone, a name or a policy detail wrong, and they can invent a flight number or a rule. Anything with a date, a number, a name or a promise needs a human check against the real source before it leaves your desk.

Watch for bias too. If you ask a tool to describe an ideal candidate for a visitor greeter or to shorten an email to someone with a non-Western name, it may carry stereotypes or mangle the name. Read for fairness, and make sure a person’s name is spelled the way they spell it.

Disclosure, policy and saying no

Find out whether your workplace has an AI policy. If it does, follow it. If it does not, ask your manager before you use AI on anything that goes to clients or leadership. You do not need to announce every spell-check, but when AI shaped a document that matters, a simple note such as “drafted with AI assistance and reviewed by me” builds trust.

Sometimes the right answer is no. If a leader asks you to put a confidential file into a public chatbot to save time, you can say, “I want to protect this information. Can we check which tool the company has approved?” That sentence protects the company, your boss and you.

Practice now

Take a real email from your work. Rewrite it with all names, numbers and private details replaced by placeholders, so it is safe to use with a free tool.
Look up whether your employer has an AI or data policy. If it has one, write down three rules that apply to you. If it does not, draft one question to ask your manager.

Use any AI assistant you have. Do not paste private, client or patient information into it.

Check yourself

1. Your boss asks you to paste an unsigned contract into a public chatbot to summarize it. What is the best response?

2. How can you still get AI help with a message about a colleague's leave?

3. Which output needs the closest human check?

Ask the AI tutor

Ask anything about this lesson, in any language. The tutor is an AI: it can be wrong, and it never needs your personal details.

Next lesson: Grow with AI: your 30-day plan

AI-drafted. Checked by an automated validator and a separate AI reviewer; a person has not read every lesson. Translations and tutor answers are AI-made. Found a mistake? Tell us. AI can be wrong. General information only, not professional advice. Not affiliated with any AI company named. Current University is not accredited.