Lesson 3 of 4 · about 5 min
Guardrails for office managers
When you finish, you can decide what never goes into an AI tool, how to check safety and legal content, and when to say no.
Stuck? Ask the AI tutorYou hold sensitive information
Office managers see staff addresses, emergency contacts, payroll questions, medical or leave notes, visitor logs, security codes, keys, alarm details, contracts and bank information. Privacy laws such as GDPR and others in many countries limit how personal data is handled.
Do not paste staff details, visitor lists, passwords, door or alarm codes, bank or card numbers, contracts or anything marked confidential into a consumer AI tool unless your employer has approved that tool for that data. Use general descriptions and placeholders like [Staff member A].
Accuracy, especially for safety and law
A tool might invent a fire-safety rule, a first-aid requirement or a lease term. It may also state a rule from another country. Requirements for fire exits, first aid, accessibility, waste and workplace safety differ by place and are updated.
Use official sources, your building manager, your insurer or a qualified adviser for safety, legal and insurance questions. Use AI to prepare questions or a clearer checklist, not to decide what is required. When people’s safety is involved, check twice.
Fairness, tone and people matters
Staff questions can touch on pay, leave, conflicts and complaints. These belong with your manager, HR or the right process. Do not use a tool to judge who is right in a dispute, or to write a warning letter without review by someone responsible for people decisions.
Check messages for tone and fairness. A notice about a shared kitchen should not sound like blaming one person. Make sure announcements are accessible, for example for staff with different languages or reading needs.
Disclosure and saying no
Follow your workplace’s AI rules. If a document is officially important, such as a policy or safety procedure, have it approved by the right person and keep a record of who checked it. Tell people when a chatbot, not a person, is answering their questions.
You can say no to a shortcut. If someone asks you to paste the staff contact list into a public chatbot to create a phone tree, you can say, “Let’s make a version with fake names, or ask IT which tool is approved.” Protecting colleagues’ information is part of the job.
Practice now
- List the kinds of information you handle that should never go into a consumer AI tool. Compare with your workplace policy or ask your manager.
- Ask a free AI tool a safety-related question for your region, then check the answer against an official source and note where it is wrong or incomplete.
Use any AI assistant you have. Do not paste private, client or patient information into it.
Check yourself
1. A colleague wants to paste the staff emergency contact list into a free chatbot to build a phone tree. What is the best response?
2. How should you treat an AI statement about required fire exits?
3. Who should review a warning letter drafted with AI?
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.
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.