Give AI Better Context: A Practical Guide for Better Work
Most disappointing AI output is a context problem, not a tool problem. Here is what to include and what to leave out.
When AI output is generic, the usual cause is a generic request. The model has no idea who you are, who is reading, what has already been decided, or what a good answer looks like in your organization. Supplying that is the highest-return habit you can build.
Five things worth including
- 01Role and situation — who you are, what the work is part of, and what has already been decided.
- 02Audience — who reads this, what they already know, and what they will do with it.
- 03Source material — the actual notes, data, policy, or draft, rather than a description of it.
- 04Constraints — length, format, tone, what must be included, what must not appear.
- 05A standard of quality — an example of good work, or a clear description of what good looks like here.
Help me write a proposal for a new reporting process.
I manage a six-person operations team. Our monthly reporting takes three days of manual work across four spreadsheets. I am proposing a consolidated process to our director, who cares about accuracy and staffing cost, not tooling. Write a one-page proposal: current state, the problem in plain terms, the proposed process, effort to implement, and what improves. No jargon, no bullet-point marketing tone. Here are my rough notes: [paste].
What not to include
- Confidential customer, employee, or financial data that your organization has not approved for AI tools.
- Credentials, keys, or anything from a system you are not allowed to export.
- Long background that has no bearing on the answer — it dilutes rather than helps.
Reusable context
You will find yourself typing the same background repeatedly. Keep a short reusable block in a note — your role, your team, your audience, and your writing preferences — and paste it in when it is relevant. Update it when your work changes.
I am an operations manager at a mid-sized logistics company. I write mostly for internal leadership. Preferences: short paragraphs, plain language, no marketing tone, state trade-offs explicitly, never invent numbers — ask me for them instead.
Iterating instead of restarting
If the first result misses, do not rewrite the request from scratch. Say what was wrong and what to keep: "The structure is right. The second section is too abstract — rewrite it using the two examples from my notes, and cut the closing paragraph." Precise correction beats a new attempt.
What to do next
- Write your reusable context block today — ten minutes, one note.
- On your next request, add audience and constraints before anything else.
- Review the result with the AI Output Review Checklist.