What AI Does Well and Badly

Models are strong at drafting, summarizing, rewriting, sorting and explaining. They are weaker at exact arithmetic, recent events (unless the tool searches), niche facts and anything that needs real-world judgment.

Quick guide

• Good: first drafts, summaries, brainstorming, formatting.
• Risky: citations, legal or medical claims, numbers.
• Poor: anything where you can't check the result.
• Good practice: a person is responsible for what gets sent.

Example (fictional)

Rewriting a messy paragraph for clarity is a good use. Asking for the current price of a product isn't. Summarizing a long email is good. Asking for the legal deadline in your state is risky. Match the task to the strength.

Common mistakes

• Using it for exact figures without checking.
• Skipping it for tasks it is actually good at.
• Sending results you could not verify.

Exercise

List ten tasks from your week. Mark each good, risky or poor for AI.