# Description: Asking for Great Results

Most weak results are not the AI being weak; they are a thin brief. The good news: a few simple habits when you describe the work make the difference between a vague answer and finished work you can use.

## Give context, not commands

Everything you type to an AI employee is a prompt: the request, plus whatever background you include with it. A terse order like "write a follow-up email" leaves it guessing about who, why, and in what voice. The same ask with context, who the email is to, what happened on the last call, and the tone you want, gives it enough to get it right.

Treat it like briefing a capable new hire. Tell it the goal, who it is for, any constraints, and what a good result looks like. The minute or two you spend describing the work up front saves the back-and-forth of fixing a generic first draft.

- Say who it is for and why it matters, not just what to make.
- Name the constraints: length, tone, format, what to avoid.
- Describe what "done well" looks like so it knows the target.

## Show an example or two

Asking with no example is zero-shot prompting: you describe the result and hope it matches your taste. Adding one or two samples of the output you want is few-shot prompting, and it steers the result far harder than more instructions ever could.

A single example carries the details that are hard to put into words: the format, the rhythm, how formal to be. Paste a past email you liked, a row from the report you want, or a heading style, and say "match this." It now has a target to copy instead of a description to interpret.

## Ask it to think step by step

For anything with more than one step, comparing options, planning, working through a calculation, ask it to reason before it answers. Letting it lay out the steps first is called chain-of-thought, and it reliably produces a more careful, more correct result than demanding the answer outright.

In practice this is as simple as adding "think it through step by step before you answer." It also makes the work easier to trust: you can see how it got there, catch a wrong assumption early, and point it back on track instead of guessing why the answer felt off.

## The standing brief

Beyond what you type in each message, an AI employee has a standing set of instructions that is always on in the background: its role, who it works for, the rules it follows, the tone it uses. That always-on brief is the system prompt, and it shapes every reply without you repeating yourself.

You usually set this once when you describe the job, not in every chat. Getting it right means your AI employee shows up the same way each time, so day to day you can give short asks and still get on-brand, in-context work back.

## Key takeaways

- A prompt is your request plus its context; rich context beats a terse order.
- One or two examples (few-shot) steer the output harder than more instructions.
- For multi-step work, asking it to reason first (chain-of-thought) improves the answer.
- The system prompt is the always-on brief that sets its role and rules, set once.
- Better results usually come from a better description, not a "smarter" AI.

## Continue learning

- [Working With AI Employees](/en/free-ai-courses/working-with-ai-employees) — View the complete course.
- [Next: How It Thinks: Context, Tokens, Reasoning](/en/free-ai-courses/working-with-ai-employees/how-it-thinks)