Prompting, checking and staying safe
Write better prompts
A prompt is the instruction you give the AI. Vague prompts get vague answers. Strong prompts usually include four parts:
- Context
- Who you are and what's going on. "I'm a project coordinator at a small nonprofit."
- Task
- Exactly what you want. "Write an email asking volunteers to confirm their shifts."
- Format
- How the answer should look. "Under 120 words, with a clear subject line."
- Constraints
- Tone, audience and anything to include or avoid. "Friendly but professional. Mention the Friday deadline."
Then refine. Treat the first answer as a draft and ask for changes: "Make it shorter," "Use simpler words" or "Give me three other options."
Always check the output
AI tools can produce hallucinations: statements that sound right but are false or made up, including invented statistics, quotes and sources. Before you use AI-generated work:
- Verify facts, numbers, names and dates against reliable sources.
- Test formulas and code on data where you already know the answer.
- Read the whole response and make sure it says what you mean.
Protect private information
Don't paste confidential information into an AI tool unless your organization has approved that tool for it. That includes customer details, health or financial information, passwords and unreleased plans. Many workplaces have an AI policy, so read yours and follow it.
Use your judgment
AI can repeat biases found in the text it learned from, and it doesn't know your workplace the way you do. Review its suggestions for fairness and accuracy. If you submit work that AI helped you draft, you are responsible for it.