Writing good prompts

Most bad answers come from thin prompts. The model is rarely what's holding you back. Type one line and the AI has to guess at everything you left out, so what comes back is generic.

Role: who it should be, like "you're a recruiter screening for a sales role." Context: what you're doing, who it's for, and any constraints. The ask: the exact output and format you want, like "five bullets, plain language, no jargon." An example: paste something you liked so it can copy the shape.

Here's the difference in practice. "Write a job post for a salesperson" gets you beige boilerplate. Compare that to "You write for a 12-person B2B startup. Write a job post for our first enterprise sales rep. Keep it punchy, skip the corporate cliches, and use short sections for the role, who we are, and how to apply. Match the tone of this post I liked: [paste]." Same model, and now it's something you'd send. The habit that works: talk to it like you're briefing a new hire on their first day.

Use this whenever the first answer comes back flat or off-target. Adding context beats rephrasing the same thin ask. It matters most on real work with a specific reader or a bar to clear, so writing, analysis, planning, anything you'd hand to a person.

Skip the ceremony for quick lookups like a date or a definition. And don't try to land it in one shot. Send a rough version, read what comes back, then steer it, maybe shorter, maybe less formal. The AI keeps what you told it, so each round gets closer.