Before I Ask AI, I Give It Context

A request such as “write me an update” leaves an awful lot unexplained.

An update for whom? About what? What does the reader already know? Do they need to make a decision, understand a problem or simply know where something stands?

Then there is my own role. The responsibilities I hold affect what I need to communicate. The audience affects the level of detail. My tone of voice affects whether the result feels like something I would actually send.

Those are the things I want AI to have in view. A document can be beautifully structured and still be completely wrong for the occasion. We have all met a paragraph wearing a suit to say absolutely nothing.

My context brief gives the tool a clearer picture of the person asking and the environment in which the work needs to be useful.

The useful parts of my own brief are my role and responsibilities, the people around my work, and the way I need to communicate with them.

For a reader making their own version, I would begin with a short description of your role. Explain what you are responsible for, then describe the audiences you regularly communicate with and the kind of information they need from you.

Next, describe your writing preferences in concrete terms. “Professional” can mean very different things to different people. You could specify that you want the main point early, a clear explanation of the implications and a direct request where a decision is needed.

Those are suggested instructions to adapt, rather than a transcript of my own master prompt. The point is to make your preferences usable.

Keep the context relevant to the work. You can often describe people by their roles rather than their names. Use only the information and AI environment your organisation permits.

A background description cannot tell AI what you need today.

For each request, there is still a purpose, an audience and source material. A useful task brief says what the output should accomplish and gives the information needed to draft it.

Here is a fictional example of the difference.

A broad request: “Make a presentation about the project.”

A more useful request: “Draft an outline for a short project update for a management audience. They need to decide whether to approve the next phase. Use only the project notes I provide. Put the decision needed near the beginning, followed by current progress, unresolved issues and the proposed next step. Mark any missing evidence rather than filling it in.”

That example isn’t a project from my job or a prompt I have claimed to test. It illustrates how much of the work sits in explaining the purpose before asking for the presentation.

It also makes the draft easier to review: you have something specific to check it against.

I often dictate what I want rather than typing the whole thing out. AI then helps me work it into a draft.

You don’t have to make your first explanation sound like a finished document. If speaking is easier, explain who the output is for, what you are trying to say and what you need them to do with it.

Then give the source material and ask for the form you need. That might be an outline, a draft update or material for slides.

The thinking is still yours. You are giving yourself another way to get it onto the page.

The following is a starting point created for this article. It is not my private work prompt, and it has not been tested across every AI product.

Fill in the parts that are relevant, then add a specific task below it. You can also save a copy somewhere you can find it again.

You can shorten this considerably for a small request. A quick piece of wording doesn’t need the same briefing as material for a decision.

I use the phrase “master prompt” to describe the context I provide. It isn’t a universal product feature. Different tools handle instructions and conversation context differently, so don’t assume that saving a brief in one place means every future request will use it.

For product-specific prompting guidance, Microsoft publishes an introduction to writing Copilot prompts, listed below. The template here is ordinary text you can adapt; it doesn’t depend on a particular settings menu.

Whatever the tool, I still assess the output myself. I check whether it is correct, whether it says what I intended and whether it is appropriate to share.

A better brief gives me a better starting point. The responsibility for the finished work stays with me.

I automate whatever I can, and I find these tools very helpful. There is still plenty left unfinished when I am busy.

My context brief hasn’t removed judgement, responsibility or the need to decide what matters. It helps me explain what I need so that the drafting has a clearer direction.

That is where I would start if you are trying to make AI more useful in your own working life: explain the role, explain the audience and explain the task. Then read the result as carefully as something you intend to put your name to.