There's no magic prompt. A two-minute experiment shows why the context you give AI matters far more than how you word the request.
Try this experiment today. It takes two minutes, and it changes how most people use AI.
Step one, ask with no details
Open Claude or ChatGPT and ask it to write a follow-up email to a customer. That's it. No details.
Read what comes back. Generic, right? It could've been written for anyone, because it was.
Step two, ask like you're briefing a new employee
Now ask again, but brief it like a new hire doing the job for the first time. Tell it four things.
- Who the customer is
- What happened on the job
- What you want them to do next
- How you actually talk to customers
Now read that one.
Why does context matter more than the prompt?
That jump in quality wasn't the AI getting smarter between your two tries. It was the context.
The AI can't see your inbox or remember your customer unless you give it that. Without context, it fills the gaps with the most average version of the email it can write. With context, it writes your email.
There's no magic prompt. There's just what you'd tell a new hire doing the job for the first time.
Anthropic, the company that makes Claude, puts it as a simple test in its own prompting guide. Show your request to a coworker who knows nothing about the task and see if they could do it. If they'd be lost, the AI is too.
What should a good brief include?
Think about what you'd tell a person, and most good briefs end up covering the same handful of things.
- What the finished piece needs to accomplish
- Who it's for and what they already know
- What happened, in your own words, with the email thread or your notes pasted in if you have them
- What done looks like, like the length, the format, and anything it has to include
- How you sound, by pasting in something you've written before and asking it to match it
You don't need all five every time. A quick reply needs a sentence of background. A proposal needs a lot more.
How much time does this actually save?
More than most people expect. In a study published in Science, professionals using ChatGPT on everyday writing tasks finished 40% faster, and the quality of their work went up 18%.
That was with no special setup. In my experience, a good brief closes the gap even more.
Do I need to learn prompt tricks?
No. Talk to it like a capable coworker. Clear context beats clever wording every time. That's the whole skill.
What do I do with a first draft I don't love?
Push back on it, the same way you would with a person. Tell it what to change in plain language, like "make it warmer, cut the middle paragraph, and add a line about the timeline." It turns a draft around in seconds.
The first answer is a starting point. Most of the good work happens in the second and third round.
If this way of thinking clicks for you, the bigger idea behind it is in AI Is a Coworker, Not Software.