How to Write a "Who We Are" File Your AI Can Actually Use

By GetGood · July 21, 2026 · 6-minute read

Why one file changes everything

Out of the box, an AI tool knows a great deal about the world and nothing about your business. Ask it to write a reply to a customer and it produces something that could belong to any company on earth — polite, generic, subtly not you. That's not a flaw in the tool; it's missing context. The fix isn't a cleverer prompt every single time. It's a standing document the AI reads before it does anything, so your business's facts and voice are always in the room. We've made the broader case that your AI is only as good as its office — the "Who We Are" file is the first and most important thing you hang on the wall of that office.

Think of it the way you'd think about a new hire's first day. You wouldn't hand them the phone and hope. You'd tell them what you do, who your customers are, how you like things said, and the handful of rules that matter most. This file is exactly that briefing — written once, so you never have to give it again, and read by the AI every single time.

The seven sections (write one or two lines each)

Keep every section short and concrete. The goal is something the AI can act on, not a brochure. Here's what goes in each.

1. Who we are and what we do

One or two plain sentences. "We're a family-owned HVAC company in Orlando. We install and service residential air conditioning, and we've done it for 40 years." Name, location, what you actually sell, and how long you've been at it. This is the anchor everything else hangs from.

2. Who we serve

Describe your real customer in a sentence or two — who they are, what they usually want, and how they tend to talk. "Mostly homeowners aged 45 and up, often stressed because their AC just failed, who want a straight answer and a firm price, not a sales pitch." When the AI knows who's on the other end, it stops writing for a committee.

3. How we sound

This is your voice, in plain instructions. "Warm, direct, no jargon. We talk like a trusted neighbor, not a corporation. Short sentences. We never oversell." If you have three or four sentences you've actually written to customers that feel like you, paste them in as examples — real samples teach voice better than any adjective. That's the heart of training AI on your voice, and this file is where it lives.

4. Our firm rules

The handful of things that are always true, no exceptions. "We always quote a firm price before starting work. We never diagnose over the phone. We don't work Sundays. We always confirm an appointment in writing." Five clear rules the AI can follow beat a page of philosophy it can't. These are the lines that keep an AI draft from promising something you'd never promise.

5. Standard answers

The dozen questions you answer every week, with your real answers. "Do you offer financing? Yes — through X, ask us for details. What's your service area? The greater Orlando metro. Are estimates free? Yes, for installs; service calls have a diagnostic fee." This section alone can turn an AI into a decent first responder for routine questions, because it's answering with your facts instead of guessing.

6. What to do when unsure

Tell the AI its escape hatch. "If you don't know, say so and offer to have a person follow up — never invent a price, a policy, or an availability." This one line prevents the most damaging AI mistake there is: answering a question it should have handed to you, confidently and wrongly. It pairs with the draft-and-approve rule — the AI drafts, you approve, and when it's unsure it flags rather than fabricates.

7. Things to never say

A short "do not" list. "Never guarantee a repair will last. Never quote a price we haven't confirmed. Never claim we're licensed for work we don't do. Never share a customer's information." Naming the forbidden lines directly is more reliable than hoping the AI infers them.

What to leave out

Two things. First, anything private or sensitive — passwords, customer records, financials, anything you'd regret being repeated. The safe test: this file should read like something you'd hand a sharp new hire on day one. Enough to represent you well; nothing that would be a problem if a stranger saw it. Second, leave out vague mission-statement language. "We value excellence and integrity" gives the AI nothing to do. "We always call back within one business day" gives it a rule it can follow. Every line should be something the AI can act on.

Where to keep it (and why format barely matters)

A plain-text or Markdown file — a .txt or .md — is ideal, because every AI tool reads it cleanly and you can edit it in any basic text editor. A Word or Google doc works fine too; the content matters far more than the format. What matters is two things: keep it somewhere your AI setup can actually read from, and keep the master copy on a machine you control, not locked inside a tool you'd lose if you stopped paying. This is the "brief" in the one-folder, one-brief, one-automation starter setup — the smallest AI office that's genuinely useful, and this file is its beating heart.

The file gets smarter every time you correct it

Don't try to make it perfect on day one. You can't, and you don't need to. Write the seven sections in an afternoon, start using your AI with it, and then do one small thing: whenever the AI writes something that makes you wince, ask whether it's a one-off or a missing rule. If it's a missing rule, add the line. Over a few weeks the file quietly absorbs how your business actually works, and the AI's output stops needing edits. That's the whole method behind a digital twin of your business — a written record of settled decisions that grows from real corrections, not from a blank-page attempt to be complete. Start rough, correct as you go, and it compounds.