SESSION 01 · 2026-09-05 · 58:17 edited replay
Better prompts. More useful AI.
Stop asking AI to guess what you mean
In our first class, I showed why a short, vague request usually gives you a vague answer. Bring your own facts, let the AI ask what it needs to know, and check what comes back before you use it.
What you will learn
- Use CRIT: Context, Role, Interview, Task. We also discussed asking for sources when facts matter.
- Work through one real task instead of collecting clever prompts.
- Spot the parts of an answer that sound right but need checking.
Follow the exercise
- Choose one real task. Pick something you actually need to do: a job application, a business idea or a work problem. Say what you need at the end and who will use it. If a CV belongs to somebody else, get permission before sharing it.
- Bring the ingredients. This is the oven, not the microwave. Give it your goal, the facts, the constraints and an example if you have one. If you do not know something, say so. Do not let the AI quietly make it up.
- Use CRIT. Give the context. Tell it what sort of help you want. Ask it to interview you, one question at a time. Then give the task and the format you need. Calling it an adviser does not make it a qualified professional.
- Answer and refine. Answer with your real experience. Read the draft. Tell it what is wrong, what sounds unlike you and what it has missed. A first answer is a starting point, not the finished work.
- Check before acting. Look at every claim you might repeat to somebody else. In an application, check it against your CV and the job description. In a business plan, check costs and rules yourself. Ask for sources, open them and make sure they actually support the claim.
Your practice task
Try CRIT on one task of your own. Save your prompt, answer the questions, and keep the draft. Write down three things you corrected or checked before you would use it.