Getting Great Answers
You don't need "prompt engineering" jargon to get great results. You need to give the model what any human helper would need. Remember four simple moves β Context, Specifics, Format, Examples:
- Context β who it's for and why. "I'm a nurse writing to a worried patient's familyβ¦" tells the model the tone and stakes.
- Specifics β the details that constrain the answer. "β¦about 150 words, warm but professional, mention we'll call with updates."
- Format β how you want it shaped. "Give me a bulleted checklist," or "a table," or "three subject-line options."
- Examples β show one if you have it. "Match the style of this past email: β¦"
Compare:
β "Write an email."
β "Write a short, reassuring email to a customer whose order is delayed two days. Apologize once, give the new date, offer 10% off their next order. Friendly, not corporate. Under 120 words."
The second one will be usable on the first try. A great rule of thumb: if a new human assistant would have to ask you a follow-up question, add that detail to your request. And if you're not sure what details matter, you can literally ask: "What do you need to know from me to do this well?"