How to Brief AI Like a Senior Hire

You ask AI for something, get a flat answer, and decide the tool is not that good. Most of the time the tool is fine and the brief was thin. A new senior hire given the same one line would also come back with something generic.

The fix is to brief AI the way you would brief a capable person who has never worked with you. Not longer, just complete. This guide gives you a five part brief you can reuse for almost any job, then walks through one example from start to finish.

A good brief answers five questions: what is the job, who is it for, what does good look like, what must it avoid, and what should it ask you before it starts.

Why one line briefs fail

Think about what a person needs before they can do good work. They need to know why the work exists. They need to know who will read it. They need an example of the standard. And they need to know the traps. If you skip those, they fill the gaps with the most average guess they can make.

AI does the same. When it has no audience, it writes for everyone. When it has no standard, it writes what most pages look like. The result reads fine and says nothing, which is the "generic" feeling people complain about.

The five part brief

Use these five parts, in this order. Each one removes a kind of guesswork.

1. The jobOne sentence on what you want made and why it exists. "A welcome email for people who just downloaded our checklist, to get them to book a call."
2. The readerWho is on the other end, what they already know, and how they feel. "Busy dentists, sceptical of marketing, reading on a phone."
3. What good looks likePaste one example you like, or describe the standard in plain words. Length, tone, structure.
4. What to avoidThe traps. Phrases you hate, claims you cannot back, topics that are off limits, anything you got wrong last time.
5. Questions firstTell it to ask you what it needs before it writes anything, and to stop there.

The last part matters more than it looks. People expect AI to start writing straight away. Asking it to question you first catches the gaps in your own thinking while they are still cheap to fix.

A worked example

Say you want a short email announcing a lesson. Here is the weak version and the brief version.

The weak version: "Write an email about my new lesson on setting up AI."

You will get a friendly, forgettable email that could have come from any company.

The brief version:

I want a short email announcing Lesson 1.1, which shows people how to set up AI so it does real work on their computer. The reader is a small business owner who has tried chat tools but never got past copying and pasting. They are curious and a little tired of hype. A good email is under 150 words, in plain words, one idea, one button. Here is an email of mine I like: [paste]. Avoid the words "revolutionary" and "game changer", and do not promise a specific amount of time saved. Before you write anything, ask me up to five questions you need answered.

The second version takes two extra minutes to write. It saves the three rounds of "make it shorter, make it less salesy, not like that" you would otherwise go through.

Keep your brief and reuse it

You should not retype the brief every time. Keep the parts that do not change, such as your reader, your tone and your list of things to avoid, in one place and paste or load them at the start. The parts that change each time are only the job and the example.

This is also where tools that remember become useful. Anthropic's docs note that each Claude Code session begins with a fresh context window, and that a CLAUDE.md file is one way to carry standing instructions across sessions. The full walk through of setting that up is in Lesson 1.1, which shows how to give AI a working setup instead of a blank chat.

How to check the first draft

When the draft comes back, check it against your own brief before editing it by hand. Ask three things. Does it do the job you named? Is it written for the reader you described? Did it avoid every item on your list?

If it misses, do not just fix the text. Fix the brief, then run it again. That way the next job starts better than this one did.

A short template to copy

Paste this at the start of a new job and fill in the gaps.

Job: [one sentence on what and why]. Reader: [who, what they know, how they feel]. Good looks like: [example or standard]. Avoid: [traps]. Before you start, ask me up to five questions, then wait.

Keep it somewhere you can reach in one click, and adjust it after each job. After a few rounds you will have a brief that fits how you work, and the first draft will stop feeling like a gamble.

Questions people ask

How long should an AI brief be?

Long enough to answer the five questions and no longer. For most jobs that is one short paragraph plus an example.

Do I need an example every time?

It helps more than almost anything else. If you do not have one, describe the standard in plain words, such as length, tone and structure.

Should I tell AI to ask questions first?

Yes for any job that matters. It exposes the gaps in your thinking before it spends effort on a draft.

What if the first draft is still wrong?

Treat it as feedback on the brief. Change the part that was unclear and run it again instead of patching the text by hand.

Where to go next

If you want AI that keeps your standing brief for you, start with Lesson 1.1. It shows the setup step by step, in plain words.