# AI at work: good at, bad at · AI in Bed

Beginner level · 6 min · https://www.aiinbed.com/learn/beginner/ai-at-work

*Where today's AI genuinely saves time, where it quietly makes things worse, and how to tell the difference.*

AI is not good or bad at "work". It is very good at some tasks inside a job and poor at others. The skill is splitting your work into tasks and handing over the right ones.

### Good at: first drafts

Emails, outlines, summaries, meeting notes, slide structure. A first draft in seconds that you then fix is often faster than a blank page.

### Good at: translating between forms

Turning notes into a table, jargon into plain English, a long report into five bullets, English into Spanish, a description into a spreadsheet formula.

### Bad at: facts you can't check

Ask for statistics, citations or legal details and it may invent convincing ones. Use AI to find and explain sources, then read the sources yourself.

### Bad at: knowing your context

It doesn't know your company's politics, last week's decision or what your boss really meant. That judgement is still yours.

**Quick check:** Which task is the best fit to hand to an AI assistant?

A. Deciding which employee to promote
B. Turning your messy meeting notes into a tidy summary you then check
C. Quoting the exact figure from a law you haven't looked up
D. Signing a contract on your behalf

Answer: B. It's a transformation task with a human check at the end. The others need judgement, accountability or verified facts.

> A useful rule: use AI where checking the answer is much faster than producing it yourself.

### Mind what you paste

Many workplaces have rules about what you may share with AI tools. Customer data, passwords and confidential plans usually shouldn't go into a public chatbot. Check your company's policy.

**Quick check:** Why is "easy to check" a good test for handing a task to AI?

A. Because AI is never wrong on easy tasks
B. Because you can catch its mistakes cheaply, so you keep the time saved
C. Because hard tasks are illegal for AI
D. Because checking makes the AI learn

Answer: B. AI makes mistakes. When checking is quick you catch them and still come out ahead. When checking is as hard as doing, you've saved nothing.

**Key takeaways**

- Split jobs into tasks: drafting and reformatting suit AI; judgement and unchecked facts don't.
- Use AI where checking is faster than doing.
- Follow your workplace's rules on what data you share.

**Go deeper (free sources)**

- [AI for Everyone](https://www.coursera.org/learn/ai-for-everyone) by Andrew Ng, DeepLearning.AI
- [Elements of AI](https://www.elementsofai.com/) by University of Helsinki & MinnaLearn