Two people ask AI for the same kind of help. One gets a generic answer and one gets a sharp one, and the difference is almost never the prompt wording, the magic phrases, or the tool. It’s context: what each of them told it about the actual situation before asking for anything.
This sounds too simple to be the answer, so almost everyone skips past it in search of something more technical. Meanwhile the people getting remarkable results are doing something that looks embarrassingly basic: they write three paragraphs of background before their one-line request.
Why context is hard (it’s not laziness)
Everything about your situation is so obvious to you that you can’t see you know it. You’re steeped in your company, your customer, your constraints, and familiarity has made all of it invisible. So when you type “write a proposal for a new client,” you genuinely don’t notice the forty things you know that AI doesn’t: who the client is, what they rejected last time, why the deadline is fake, which of your services actually fits, what tone this particular buyer trusts.
Psychologists call this the curse of knowledge. Teachers fight it, senior engineers fight it, and now everyone with a chat window fights it. Articulating what you know, to something that knows nothing about you, turns out to be a demanding communication skill. Nobody taught it, because until recently nobody needed it daily.
How to give AI better context
Give AI five things before your request: your situation, your goal beyond the document, the audience for the output, your constraints, and what good looks like. Plain sentences are enough, with no special formatting. Then make one request and refine from there. The difference between a generic draft and a sharp one is almost always this background, not the prompt wording.
The situation: who you are, what’s happening, and the one-paragraph history that led here. The goal: what you’re actually trying to achieve, beyond this document (“win back a client who went quiet,” not “write an email”). The audience: who will read or use the output, and what they care about. The constraints: length, tone, deadline, budget, the things you’ve already tried or ruled out, anything that must or must not appear.
And what good looks like: a description of the end product you’d be delighted with, or better, an example of past work you consider excellent.
Then make one request, and when the draft comes back, tell it what’s close and what’s off; the back-and-forth is where the tailoring happens.
If five things feel like too many to remember, remember one sentence instead: tell it everything you’d tell a bright new hire on their first morning, before you’d trust them with this task.
Briefing AI is the same skill as delegating
Situation, goal, audience, constraints, what good looks like: that’s a delegation briefing. It’s what capable leaders have always given people before handing off work that matters, and the reason some managers’ teams flounder is the same reason their AI output is generic – the context stayed in their head.
So one practice improves two skills. Every hour spent learning to put what you know into words makes you better with every AI tool at once, and, along the way, better at handing work to humans, which is the more valuable skill of the two. I work with people on this, because it sits at the junction of communication, leadership, and thinking clearly, and because the benefit keeps growing as AI improves.
The fix, then, rarely involves a new tool. The forty things you know and haven’t said are the difference between generic and sharp, with AI and with people, and writing them down is a skill worth practising for the rest of your career.