Hands-on: getting great answers

The same model gives a far better answer when you ask well.

You don't need tricks or "magic words." You need to give the model what a new colleague would need: context, a clear task, and the shape of the answer you want.

See the difference

Same goal, two prompts

Vague prompt
Clear prompt
You
write something about our product launch
The model
Here is a piece about a product launch: "We are excited to announce our new product. It has many great features and we think you will love it. Stay tuned for more details…" — generic, could be about anything, you'll be editing for ages.
You
You're our marketing lead. Write a 4-sentence LinkedIn post announcing the launch of "InventoryIQ," a tool that helps small shops predict stock shortages. Audience: shop owners. Warm, plain, no buzzwords. End with a question to invite comments.
The model
"Running out of your best-seller at the worst moment? We built InventoryIQ for exactly that. It learns your sales patterns and warns you before you run low — so you reorder just in time, not too late. What's the one product you can never afford to run out of?" — ready to post.

The 6 habits that do 90% of the work

1

Give a role

"You are a patient tax advisor…" sets the tone, vocabulary, and depth instantly.

2

Add context

Who's it for, what's the goal, any background. It can't read your mind — only your words.

3

Say the format

"In 5 bullet points," "as a table," "under 100 words." You'll get exactly that shape.

4

Show an example

Paste one example of what "good" looks like. One sample beats a paragraph of instructions.

5

Iterate, don't restart

"Make it shorter," "warmer," "add a stat." It's a conversation — refine in steps.

6

Ask it to think

"Think step by step, then answer." Helps on tricky tasks. (Newer "reasoning" models already do this on their own.)

The golden rule: if a smart new hire couldn't do the task from your instructions alone, the model can't either. Vague in, vague out. Specific in, useful out.

Prompt it well and it gives great answers — but notice it still only ever talks back. The last step is wiring that talent up so it can actually do things for you.