Last updated: December 2025
Your company probably has an AI policy. You probably haven’t read it. And even if you have, it likely doesn’t cover the gray areas where most AI use actually happens — the quick ChatGPT query about a client project, the Claude-assisted email to your boss, the Copilot-generated code that’s now in production.
AI at work is a minefield of productivity gains and career risks. Here’s how to navigate it.
The Rules Nobody Tells You
Rule 1: Your Company’s AI Policy Is the Floor, Not the Ceiling
Most corporate AI policies say something like “don’t upload confidential data to AI tools.” That’s the minimum. The unwritten rules are stricter:
- Don’t let AI make you look incompetent. If you submit AI-generated work that’s clearly wrong, you own that mistake. “The AI produced it” is not a defense; it is an admission that you did not review your own work.
- Don’t let AI make you look lazy. Submitting obviously AI-generated text (generic, formulaic, full of “explore” and “landscape”) signals that you didn’t put in effort. Edit AI output until it sounds like you.
- Don’t use AI for things that require your judgment. Performance reviews, hiring decisions, legal opinions, medical advice — these require human expertise and accountability. AI can help you prepare, but the final output must reflect your professional judgment.
Rule 2: Disclosure Depends on Context
When should you tell people you used AI?
Always disclose:
- Published content (articles, reports, white papers)
- Client deliverables where originality is expected
- Academic work
- Legal or compliance documents
- Anything where someone is paying for your expertise specifically
Usually don’t need to disclose:
- Internal emails and messages (AI-assisted writing is like using spell check)
- Code (AI-assisted coding is standard practice)
- Research and brainstorming (AI as a thinking tool)
- Formatting and editing (AI as a productivity tool)
Gray area (use judgment):
- Presentations to leadership
- Client communications
- Creative work
- Strategic recommendations
The test: Would the recipient’s trust in you change if they knew AI was involved? If yes, disclose.
Rule 3: Never Upload What You Can’t Afford to Leak
This is the rule that gets people fired. Before pasting anything into an AI tool, ask:
- Is this covered by NDA?
- Does it contain customer data?
- Is it proprietary code or trade secrets?
- Would it be a problem if this appeared in someone else’s AI output?
If the answer to any of these is yes, don’t paste it. Use anonymized versions, or use a tool with enterprise data protection (like your company’s approved AI platform).
Real examples of people getting fired:
- Samsung engineers who pasted proprietary source code into ChatGPT
- A lawyer who submitted AI-generated legal briefs with fabricated case citations
- A marketing manager who uploaded a competitor analysis containing confidential client data
How to Use AI Well at Work
For Writing
Do: Use AI to generate first drafts, overcome writer’s block, check tone, and suggest edits. Then rewrite in your voice.
Don’t: Submit raw AI output. Ever. Even if it’s good, it’s not yours until you’ve made it yours.
Pro tip: Use Claude for first drafts (best writing quality), then edit heavily. The final product should be 60% your words, 40% AI-assisted structure. Nobody can tell the difference, and the quality is higher than either you or AI alone.
For Coding
Do: Use Copilot or Cursor for autocomplete, boilerplate, test generation, and debugging assistance. Review every line before committing.
Don’t: Commit AI-generated code without understanding it. If you can’t explain what the code does and why, don’t ship it.
Pro tip: AI-generated code often has subtle bugs — wrong edge case handling, security vulnerabilities, or performance issues. Use AI to write the first version, then review it with the same rigor you’d apply to a junior developer’s PR.
For Research
Do: Use Perplexity or ChatGPT to quickly understand topics, find sources, and synthesize information. Verify key facts independently.
Don’t: Cite AI-generated information without checking the sources. AI tools hallucinate — they present false information with complete confidence.
Pro tip: Use Perplexity (which provides source citations) instead of ChatGPT (which doesn’t) for research. Then click through to the actual sources to verify.
For Email
Do: Use AI to draft routine emails, suggest more professional phrasing, and check tone before sending sensitive messages.
Don’t: Use AI for emails that require genuine empathy or personal connection. “the assessment is sorry about your loss” should not be AI-generated.
Pro tip: For important emails, write the first draft yourself (to capture your actual thoughts), then ask Claude to “make this more professional/concise/diplomatic.” This preserves your intent while improving the delivery.
For Presentations
Do: Use AI to structure presentations, generate slide outlines, create speaker notes, and suggest data visualizations.
Don’t: Let AI determine your narrative or conclusions. The story you tell should reflect your analysis and judgment, not AI’s generic framework.
Pro tip: Use ChatGPT to generate 10 possible angles for your presentation, pick the best one, then develop it yourself. AI is great at brainstorming options; you’re great at choosing the right one.
The Skills That Matter More Now
AI doesn’t replace professional skills — it shifts which skills matter most.
Less Valuable Now
- Fast typing and formatting. AI handles this.
- First-draft writing. AI generates adequate first drafts instantly.
- Boilerplate code. AI writes this better and faster.
- Basic research. AI synthesizes information from multiple sources in seconds.
- Rote analysis. AI processes data faster than humans.
More Valuable Now
- Judgment. Knowing which AI output to use and which to reject.
- Editing. Turning adequate AI output into excellent final product.
- Prompt engineering. Getting better results from AI tools (this is a real skill).
- Critical thinking. Evaluating AI-generated analysis for errors and bias.
- Domain expertise. Understanding your field deeply enough to catch AI mistakes.
- Communication. Explaining complex ideas to humans (AI can’t do this in meetings).
- Relationship building. Trust, empathy, and collaboration remain deeply human.
Common Mistakes
Mistake 1: Using AI as a Crutch Instead of a Tool
If you can’t do the work without AI, you can’t evaluate whether AI did it correctly. Use AI to be faster and better at things you already know how to do, not to fake competence in areas you don’t understand.
Mistake 2: Over-Relying on One Tool
ChatGPT is not the answer to everything. Use the right tool for the right task. Claude for writing, Perplexity for research, Cursor for coding. Specialization matters.
Mistake 3: Not Keeping Up
AI tools improve monthly. The workflow that was optimal three months ago might be outdated. Spend 30 minutes per month trying new tools and features. Read reviews (like the ones on this site) to stay current.
Mistake 4: Being Secretive About AI Use
The stigma around AI use is fading fast. Being open about how you use AI (while following your company’s policy) positions you as tech-savvy and efficient, not lazy. The people who pretend they don’t use AI are the ones who look out of touch.
The Bottom Line
AI at work is like any powerful tool — it amplifies both competence and incompetence. Used well, it makes you faster, more thorough, and more effective. Used poorly, it produces mediocre work, creates security risks, and erodes trust.
The professionals who thrive with AI are the ones who use it as a force multiplier for skills they already have, not a replacement for skills they lack. Be that person.
Related guide: how to use AI at work.
Related guide: how to use AI at work.
Related guide: how to use AI at work.
Build your professional AI toolkit: