Quick Answer
Company context for AI is the background an AI needs to work for your business: who you are, who you serve, what you offer, the words you use, how you lay out work and how you sound. Without it, AI produces generic work that people spend time correcting. With it, every task starts from the same accurate picture of your company.
What Company Context Is
Think about a new starter's induction. Before anyone hands them a job, they learn who the company is, who its clients are, what it sells, the names of things, how documents are laid out and how the company sounds in writing. Company context is that induction, written down for your AI.
It usually covers:
- Who you are: your name, what you do, where you work and how big you are.
- Who you serve: your clients, sectors and the problems you solve for them.
- What you offer: your services or products, how they're packaged and how you price.
- Your terminology: the words you use and the ones you don't, product names and project codes.
- How work is laid out: document structures, date and currency formats, spelling and where things are saved.
- How you sound: your tone of voice, with a few examples.
- What to avoid: claims you never make, competitors you don't name, phrases you've retired.
Most of this rarely changes. That's what makes it worth setting once and reusing everywhere, instead of retyping it into every request.
Why AI Gets Your Business Wrong
An AI model knows a great deal about the world in general and nothing about your company in particular. Unless someone tells it, it fills the gaps with plausible defaults: generic packages, invented prices, the wrong spelling and a tone that sounds like everyone else. The open Agent Skills project makes the same point: AI agents are capable, but often lack the context they need to do real work reliably (Agent Skills overview).
Teams usually patch this person by person. One colleague keeps a long block of custom instructions. Another pastes in the style guide each time. A third has a project set up in one tool that nobody else can see. Each copy ages on its own, and none of them reaches the AI tool someone else prefers. When the business updates its pricing, half the copies still carry the old figures.
The result is work that varies by who asked and which tool they used. People spend the time AI was meant to save fixing names, terms and formats.
How Company Context Works With Skills and Knowledge
Company context is one of four parts that together produce work you can use straight away:
| Part | In plain words | The new-starter version | Without it |
|---|---|---|---|
| Company context | Who it's working for | The induction | Generic work in the wrong tone and terms |
| Knowledge base | How to do it, in detail | The SOPs, templates, guidelines and scripts | It makes up the detail or guesses the format |
| Skill | What to do | The job card for one task | A different result every time |
| Harness | What brings them together | The manager who hands over the induction, the right manuals and the job card before every job | Each part sits somewhere different, and the AI starts without the full picture |
The knowledge base and the harness each have their own guide. Context sits under every skill. A proposal skill doesn't need to explain who your company is, because the context already has. That keeps each skill short and focused on its job. The roles can overlap: a company profile in your knowledge base can feed your context, and that's fine. What matters is that the AI gets the right picture each time.
Worked Example: The Same Proposal, Twice
Larchwood Digital, a fictional web agency, asks its AI for a proposal for Harbour Foods, a fictional food distributor. The proposal skill is identical both times. Only the company context changes.
Without company context, the draft opens with "[Your company] is a leading provider of digital solutions". It offers three made-up packages, quotes hourly rates Larchwood doesn't charge, calls the stages "sprints" and uses US spelling.
With company context, the draft uses Larchwood's name and its real three-tier offer. It describes fixed-fee phases, because that's how Larchwood prices. It calls the stages "phases", as the team always does, uses New Zealand spelling and follows the proposal layout clients already know. The pricing figures still come from the current price list in the knowledge base, not from the context.
The difference isn't the AI model. It's what the AI was given.
Try It: Write Down Your Company Context
- Write a one-page overview: who you are, who you serve and what you offer.
- List 10 to 20 terms with how you use them, including words to avoid.
- Note your formats: document layout, dates, currency and spelling.
- Describe your tone of voice in three to five rules, each with a short example.
- List the claims and phrases the AI must never use.
- Note where teams differ, for example how sales and finance name things.
- Keep it short. Link to longer documents in your knowledge base rather than pasting them in.
- Try it with one skill you already use and compare the results.
Test It: Three Cases
- A normal request. Ask for a proposal or client update. Check the company name, offer, terms, spelling and layout.
- Incomplete input. Leave your pricing model out of the context and ask for a quote. The AI should ask or point to the price list, not invent rates.
- An unrelated request. Ask a general question, such as how a tax rule works. The answer shouldn't turn into a sales pitch for your services.
Where Contexta Fits
A skill tells your AI what to do. Your knowledge base shows it how, in detail. Your company context tells it who it's for. The Contexta harness brings them together, in every AI.
In Contexta, company context is set as organisation rules: your identity, terminology, naming, formats, where work is saved and what to avoid. Each team can add its own rules on top. Brand voice profiles keep the tone consistent. Every change to a rule is kept in version history, the same as a skill, so you can compare and restore. Our advisory and implementation team can also set up a fuller company context with you, through Contexta Managed AI.
Every session starts the same way: the connected AI loads your organisation's rules, the right space and the published skill. Your knowledge base stays in your own SharePoint or Google Drive. Company, team and personal spaces give each group its own home and rules. Update a rule once and connected AIs use it the next time they're asked. Contexta is tested with Claude, ChatGPT and Cursor; see How It Works.