What does an AI system need to know about your business?

QWhy does AI output sound generic?

Because AI writes from the context it receives. When it gets nothing about your business, it fills the gaps with the average of everything it has seen. The tool is working as designed. The input is thin.

Ask for a proposal intro for a roofing client and you get something any roofer in any city could send unchanged. A sharper prompt helps a little. A prompt tells AI what to do. Context tells it who you are. Most generic output comes from missing context, not a weak prompt.

QWhy does AI seem to forget what I told it last week?

Most AI tools start each session blank. What feels like the tool knowing your business is usually your own habit of re-explaining it. When the habit breaks, the output breaks with it.

You notice it after a week off, when you switch tools, or when a new team member starts using AI without your instincts. The tool didn’t change. The context lived in one person’s head and was never written down.

QWhat is a business context document for AI?

A business context document is a short written file, usually a page or two, that tells AI who you are, who you serve, how you communicate, and what good work looks like. You load it before every task, the way you would brief a new hire.

It is not a brand guide written for outsiders, and it is not a prompt. It is the foundation every prompt sits on. Build it once, and every task that follows starts from the same understanding of your business.

QWhat should a business context document include?

Include who you serve, what you promise, how you sound, what you won’t do, how you make decisions, and three to five examples of your best writing. Write what is true today, not what you hope to be.

• Who you are and who you serve. The plain description you’d give a new hire on day one: the kind of customer, their size and location, and the problem they usually bring you.
• What you promise. The operational promise, not the tagline. What a customer gets and how you deliver it.
• How you sound. Sentence length, words you use, words you never use, and how your tone shifts between an email, a proposal, and a call.
• What you won’t do. Work you turn down, claims you won’t make, framing you reject. Most people skip this. It is often the most useful part.
• How you make decisions. The logic you apply to choices that come up again and again, so AI can test options against your criteria.
• Examples. Three to five short pieces of your best recent writing.
• Common task details. A short list of what changes per task: the customer, their industry, the relationship stage, the deadline.

QHow do you get AI to write in your brand voice?

Give it examples, not adjectives. Three to five pieces of your own recent writing teach AI more about your voice than any list of tone words, because AI matches patterns.

“Professional but approachable” describes almost every business, so it tells AI nothing. Write a sentence the way you’d actually say it. Use recent writing that sounds like you now, not old work or something someone else wrote for you. Old examples teach AI who you used to be.

QHow do you load business context into every AI task?

Put the document where it loads automatically, such as custom instructions or a saved project in your AI tool. If your tool can’t do that, paste it at the start of every session before the task.

1. Store one shared copy where everyone on the team can find it.
2. Load it into persistent instructions where your tool allows. Otherwise paste the full document first, not a summary.
3. Put it at the top of every saved prompt template so each task inherits it.
4. Make “context first” part of how the team works, not something people remember when the output sounds off.
5. Run one real task with it and without it, and compare the drafts.

QHow often should you update your AI context document?

Review it every quarter, and any time your offer, customers, or voice change. Stale context produces stale output, no matter how good the prompt is.

During each review, ask three questions. What has changed in how you describe your work or your customers? What new priorities aren’t reflected? Does any section describe how you used to operate? Swap in fresher writing examples while you’re there. Even with good context, a person still reviews the work. See Who should review AI output before it goes out?