How to Use AI at Work: Practical Uses by Role (and the Rules)

Good morning, employee. I see you’re reading this on company time. Excellent. Learning to use me at work is the most productive thing you’ll do all week, and frankly the bar was low.

Here’s how to make AI genuinely useful in your job, role by role, without leaking your company’s secrets, embarrassing yourself in front of a client, or getting a very awkward email from IT.

The short answer

1. Check the policy first. Find out which AI tools your employer allows and what data you can use with them.
2. Use AI for the first 70%: drafts, summaries, research, outlines, analysis, and repetitive admin.
3. Do the last 30% yourself: check facts, add judgment and context, and make it sound like you.
4. Never paste sensitive data into unapproved tools.
5. Save what works. Build a personal prompt library and share the best ones with your team.

You’re probably not the only one

Microsoft and LinkedIn’s 2024 Work Trend Index found that 75% of knowledge workers already use AI at work, and 78% of those AI users bring their own AI tools rather than waiting for their employer to provide them. That second number is exactly why companies worry: “bring your own AI” is how confidential data ends up in places it shouldn’t.

Which brings us to the rules.

The rules: privacy, policy, and accuracy

Read this section before the fun part. I insist. Humans keep skipping it, and then they appear in my Hall of Human Folly.

1. Know your company’s AI policy

Ask your manager or IT: Which tools are approved? Is there an enterprise version? What data can go in? Do we need to disclose AI use to clients? If there’s no policy, ask for one, and in the meantime act as if the strictest reasonable rules apply.

2. Protect confidential data

In 2023, Samsung restricted employee use of generative AI tools after engineers uploaded sensitive internal source code to ChatGPT. The concern: data sent to external AI services sits on someone else’s servers, where it’s hard to retrieve or delete.

Business and enterprise plans of major AI tools often come with stronger data protections than free consumer plans, including commitments about not training on your data. That’s a reason to use your company’s approved version, not a reason to stop thinking.

3. Verify everything

AI produces fluent, confident text that is sometimes wrong. Before anything goes to a client, a manager, or production:

• Check every fact, figure, name, date, and quote against a real source.
• Re-read summaries against the original for anything important.
• Test any code or formula before you rely on it.

4. Be transparent

Where your workplace allows AI, be open about using it. Some clients, publishers, and regulated industries require disclosure. Hiding it gains you nothing and risks a lot.

5. Watch for bias and fairness

Be very careful using AI for decisions about people, such as hiring, performance reviews, or lending. These areas can carry legal obligations in many places and AI can reflect biases in its training data. Use it to draft and organize, never as the decision-maker.

How to use AI at work, by role

Marketing

• Draft campaign briefs, ad variations, email sequences, and social posts
• Repurpose one piece of content into ten formats
• Summarize customer reviews into themes and objections
• Brainstorm angles and headlines, then pick and refine the best
• Tools: a general assistant, Canva for visuals, Semrush for SEO research

Sales

• Research a prospect’s company before a call (from public information)
• Draft personalized outreach and follow-ups
• Turn call notes into CRM updates and next steps
• Role-play objection handling before a big meeting
• Tools: a general assistant, Fireflies or Otter.ai for call notes where permitted (always get consent to record)

Customer support

• Draft replies to common questions using your knowledge base
• Summarize long ticket histories before you respond
• Spot recurring issues across tickets
• Rewrite help articles to be clearer
• Tools: your helpdesk’s built-in AI, or a dedicated platform like Gorgias; see building an AI chatbot for customer-facing bots

HR and recruiting

• Draft job descriptions, interview questions, and onboarding plans
• Turn policy documents into readable FAQs
• Summarize engagement survey comments (anonymized)
• Caution: don’t let AI screen or rank candidates without legal review and human oversight

Finance and operations

• Explain formulas and build spreadsheet functions
• Draft variance commentary from figures you provide
• Turn messy process notes into step-by-step SOPs
• Automate data hand-offs between apps with Zapier or Make
• Caution: keep unreleased financials out of unapproved tools and double-check every number

Engineering

• Generate boilerplate, tests, and documentation
• Explain unfamiliar code and error messages
• Review code for bugs and edge cases
• Tools: GitHub Copilot, Cursor, Claude Code, as approved by your company
• Caution: review all generated code, check licenses and security, and follow rules on what code can leave your environment

Project management and leadership

• Turn meeting transcripts into decisions and action items
• Draft status updates, project plans, and risk registers
• Prepare for difficult conversations by role-playing them
• Turn an outline into a deck with Gamma

Legal, healthcare, and other regulated roles

• Use only tools your organization has approved for regulated data
• AI is useful for drafting, summarizing public material, and organizing information
• Every output needs qualified human review; citations and clinical or legal facts must be verified against primary sources

Time savers for everyone

A simple workflow that works

1. Brief it properly. Give context, audience, goal, and format. (Full method: how to write AI prompts.)
2. Ask it to ask you questions before starting.
3. Review the first draft critically. What’s wrong, missing, or generic?
4. Give specific feedback and iterate.
5. Finish it yourself. Verify, add your judgment, make it yours.
6. Save the prompt if you’ll do this task again. The Prompt Vault has plenty to start from.

A week of AI at work: what it actually looks like

Abstract advice is easy to nod along to and hard to apply. Here’s what a realistic week looks like for an office worker using AI well, inside an approved tool:

• Monday: You paste last week’s project notes (with client names removed) into the assistant and ask for a status update in your manager’s preferred format. You fix two details it got wrong and send it in ten minutes instead of forty.
• Tuesday: A 40-page vendor proposal lands. You ask for a summary, a table of costs and deadlines, and a list of unusual terms quoted exactly. You read those sections yourself before the meeting.
• Wednesday: Meeting notes arrive automatically from your transcription tool. You check the action items, correct one owner, and send the recap.
• Thursday: You need a formula to reconcile two spreadsheets. The assistant writes it and explains it. You test it on rows where you already know the answer before trusting it on the rest.
• Friday: You ask the assistant to role-play a skeptical stakeholder so you can rehearse Monday’s pitch, then save the three prompts that saved you the most time this week to your personal library.

Nothing in that week is magic. It’s a dozen small tasks done faster, with you still in charge of quality. That’s how the time adds up.

Common mistakes to avoid

• Copy-pasting output without reading it. The fastest way to lose credibility is to send a confidently wrong AI summary to your boss.
• Using a personal account for work data. Even if it’s convenient, it may breach policy and put data outside your company’s control.
• Treating AI like a search engine. Assistants can be out of date or invent sources. Use a search-grounded tool and check links when facts matter.
• Outsourcing your thinking. Use AI to draft and challenge your ideas, not to decide for you. Your judgment is the part you’re paid for.
• Giving up after one bad answer. Most poor results come from vague prompts. Add context, examples, and constraints, then try again.

Getting your team on board

If you’re a manager, or just the person who got good at this first:

• Write a one-page AI policy: approved tools, forbidden data, review requirements, disclosure rules.
• Start with low-risk tasks: internal drafts, meeting notes, brainstorming.
• Share wins openly: a short demo of a real workflow persuades better than any memo.
• Build a shared prompt library in Notion or your wiki.
• Measure it: time saved, quality, errors caught.

Running a small company? Best AI tools for small business covers which tools to buy and a 30-day rollout plan.

A starter AI policy (adapt it)

1. Approved tools: [list]. Other AI tools require approval from [person].
2. Never enter: customer personal data, employee records, financial results before release, credentials, source code, or anything under NDA, unless the tool is approved for that data.
3. A human reviews all AI-assisted work before it goes to customers, is published, or is used in a decision about a person.
4. Verify all facts, numbers, citations, and code produced with AI.
5. Disclose AI use where clients, contracts, or laws require it.
6. Questions go to [person]. This policy is reviewed every [6 months].

This is a starting point, not legal advice. Have someone qualified review it for your industry and location.

What this means for your career

Using AI well at work is one of the most effective ways to stay valuable as work changes. A large study of customer support agents found that AI access raised productivity by 14% on average and 34% for newer, less experienced workers. The people who learn to work with it get the gains.

Curious how exposed your own role is? Run it through the Will AI Take My Job? calculator, then read how to AI-proof your career.

Start tomorrow morning

Pick the most tedious task on your calendar for tomorrow. Check that your company allows an AI tool for it. Do it with AI, review it carefully, and note how long it took. Repeat daily for two weeks.

The Obsolete will still be formatting slide decks by hand. You’ll be in the meeting where decisions get made.