Your AI is only as useful as what you put into it. Write the knowledge base like a briefing for a sharp new hire on day one: specific, structured, and honest about what your business actually does. Vague inputs produce vague outputs. Seven sections, before-and-after examples, and a pre-submission checklist.
Your AI knows exactly as much about your business as you've put into it. For most businesses, that's less than you'd hand a new hire on their first afternoon.
That's what this covers: how to write an AI knowledge base your AI can actually use. It's the document your AI reads as context before every conversation. Get it right and your AI sounds like you. Get it wrong and it sounds like a chatbot that was told nothing and is doing its best.
What the knowledge base actually does
The knowledge base isn't your website copy, and it isn't your brochure. It's a briefing document: the file your AI reads before it responds to anything.
Most AI agents run on a system prompt, a set of instructions and background information injected before every conversation starts. The knowledge base is the substance of that prompt. It's where you tell the AI who your business serves, what you charge, how you speak, and what happens when a conversation goes sideways.
Most knowledge bases read like terms and conditions written at 11pm on a Friday. Technically complete. Practically useless. "We provide premium AI solutions for businesses of all sizes at competitive prices" gives the AI nothing to work with. Ask it "how much does it cost?" and it answers something like "our pricing is competitive, please reach out for details."
Which is, charitably, not helpful.
The knowledge base also decides whether your AI sounds like your business or like a generic assistant briefed on nothing. It doesn't invent personality. It reflects what it's given. Feed it marketing copy and it speaks in marketing copy. Feed it real answers and it gives real answers. Only one of those is useful to a client with a question right now.
The golden rule
Write it for a sharp new employee on their first day. Not for a search engine. Not for a regulator. For a capable person with zero context on how your business actually works.
That person needs to know: what you sell, who buys it, what it costs, what you say when someone asks a specific question, and what you never promise. That's the entire job of the knowledge base.
Your AI isn't psychic. Neither are your employees, but at least they can ask a clarifying question before saying something unfortunate.
If the information isn't in the knowledge base, the AI guesses or deflects. Guessing is charming in a pub quiz. In a client conversation about your service scope, it's a different matter.
One test: could your knowledge base double as a LinkedIn company page? If yes, rewrite it. LinkedIn is optimised for impressions. Your AI knowledge base is optimised for useful answers. Those are different jobs, and writing for one while calling it the other produces a document that does neither well.
The 7 sections your knowledge base needs
1. Business identity
Who you are, what you do, who you serve, and what outcome you deliver. Three to five sentences, real specifics, not marketing language. Your target audience, the problem you solve, and the geography you cover.
2. Services and pricing
List each service with a name, a price range, a one-sentence description, and who it's for. The AI can't invent prices. If you don't give it a number, it either refuses to answer or fabricates one, and both outcomes cost you trust.
Price ranges are fine. "Between $2,500 and $4,500 depending on scope" is useful. "Contact us for a quote" isn't information. It's an instruction to the client to do the work you should have done in the knowledge base.
3. Common questions and answers
Write 5 to 10 real questions you get, with your actual answers. Not marketing copy. Answers. Marketing copy sounds good. An answer to a specific question is useful. Conflating the two produces text that does neither.
4. Tone and style
Describe how you speak. Professional or casual? First names or last? Contractions? What phrases are on-brand, and what do you never say? This is the section that makes your AI sound like your business instead of a generic assistant. Skip it and the AI defaults to a neutral corporate register that sounds nothing like you.
The Anthropic model specification is useful background here. It explains how the model synthesises instructions rather than quoting them verbatim, which is exactly why describing your tone precisely matters more than handing over sample sentences.
5. Booking and contact flow
Calendar link, phone number, email address, and any qualification questions to ask before booking. If a prospect says "I want to start," the AI needs to know the actual next step, not just "contact us." A specific next step closes the loop. "Contact us" opens it back up.
6. What you don't do
The section most businesses skip, and regret skipping. List the services you don't offer, the clients who aren't a good fit, and where you'd refer out instead. An AI that says "that's not what we do, here's who might help" earns more trust than one that vaguely tries to accommodate everyone. Boundaries are information. Give your AI the information.
7. Escalation rules
When should the AI hand off to a person? Define at least 2 concrete scenarios. "When things get complicated" isn't a rule. It's a hope. "When the contact mentions a refund, legal action, or a custom quote above $15,000" is a rule. Write rules.
Before and after: what weak and strong entries look like
Here's the same piece of information, written two ways.
Weak: "We provide premium AI solutions for businesses of all sizes. Our experienced team delivers strong results with a focus on quality and client satisfaction."
What your AI learns from this: nothing. It can't answer "who do you work with," "what do you charge," or "are we a good fit" from that paragraph. It deflects or guesses. (Confident guessing is the AI's default move when the knowledge base is thin. Employees do the same thing, they just have the decency to hesitate first.)
Strong: "We work with service businesses, typically 10 to 50 staff, in Canada and the US. We charge between $2,500 and $4,500 for a Professional engagement depending on scope. We're not the right fit for startups without existing processes to automate, or for companies looking for a quick chatbot rather than a system."
What your AI learns: who to welcome, who to redirect, and how to answer "are we a good fit" with specifics instead of a deflection.
The difference is specificity. Every adjective you swap for a number, a category, or a concrete example makes the AI a more useful representative of your business. "Premium" tells it nothing. "$2,500 to $4,500" tells it everything it needs.
How long should it be?
Target 600 to 1,500 words.
Under 400 words is usually too thin. The AI fills the gaps with best guesses, and those guesses won't always match your brand. Over 2,500 words and the document loses focus: the AI tries to weigh everything equally and ends up prioritising nothing.
The sweet spot is a document a sharp person could read in 8 minutes and walk away with a clear mental model of your business. If it takes more than 10 minutes to read comfortably, tighten it. Don't expand it.
Word count is a proxy for clarity, not quality. A 700-word knowledge base with real prices, real FAQs, and a specific tone instruction outperforms a 2,000-word document full of adjectives and mission statements. The OpenAI prompt engineering guide makes the same point from the other direction: dense, specific instructions outperform long, vague ones, regardless of which model is reading them.
When the AI still gets it wrong
If your AI keeps producing off-brand or incorrect responses after you've submitted the knowledge base, the cause is almost always the same: a missing section, a vague answer, or a tone instruction that conflicts with something else in the document.
The pattern is consistent. A business owner spends $200 a month on an AI writing tool, gets generic blog posts that sound nothing like them, and stops using it after 90 days. Not because the tool was bad. Because nobody told the AI how they speak, who their clients are, or what makes their offer different. The AI amplifies what you put in. Give it nothing and it gives you generic. Give it real information and it gives you real answers.
Most AI implementation goes wrong at exactly this step. A consultant recommends a tool, configures the integration, and leaves. The knowledge layer, documented processes, defined voice, mapped client knowledge, never gets built. That's the actual failure point. Not the technology.
The AI-Ready Business Blueprint exists to close that gap. It's 7 documented deliverables, including a structured knowledge base, built across 8 sessions. The knowledge base section alone usually takes a full session to get right, not because it's complicated, but because most founders have never been asked to write down how their business actually works, and the first draft always turns up surprises.
If you're not ready to document your business in full, the AI won't be ready to represent it accurately. That's not a reason to skip it. It's a reason to do the documentation work first. Book a call and we'll help you find the gaps.
Your pre-submission checklist
Go through this before you submit. If you can't check every box, spend 15 more minutes on the ones you can't.
- Business identity includes your target audience and the specific outcome you deliver, not just a description of what you do
- Every service has a price, or at minimum a price range (for example, "$2,500 to $4,500")
- At least 5 real FAQs with complete, specific answers, not "it depends"
- Tone section describes how to greet people, what phrases to avoid, and whether to use first names
- A "what we don't do" section exists and is honest
- Booking instructions are specific: a calendar link, email address, or phone number for qualified prospects
- Escalation triggers are defined for at least 2 concrete scenarios
- The entire document can be read comfortably in under 10 minutes
If you've got the knowledge base ready and want to put it to work in a social channel, the next step is understanding how the agent actually uses it. Start with how AI social media comment management actually works: the knowledge base is one of the first things the system reads before it responds. The same document also configures our voice agents; see how an AI voice agent actually works for that side of it.
Other things on this site that'll save you money.
- →AI Readiness Blueprint — we build this document with you across 8 sessions.
- →AI social media comment management — one place this document gets put to work immediately.
Straight answers, marked up for Google.
Can I update the knowledge base after the AI goes live?
Yes. Submit an updated version and we redeploy the updated prompt. Simple additions, a new service, an updated price, are usually quick to turn around. Full rewrites go through a new review round.
Does the AI memorise the knowledge base word for word?
No, it reads the document as context and synthesises an answer. It won't quote it verbatim. That's why precise descriptions matter more than polished sentences: the AI extracts meaning, not exact wording.
How long should an AI knowledge base be?
Target 600 to 1,500 words. Under 400 is usually too thin: the AI fills gaps with guesses. Over 2,500 words and focus gets diluted. Aim for something a sharp person could read in 8 minutes and walk away with a clear mental model of your business.
Can I include pricing I don't want publicly visible?
Yes. The knowledge base is injected into the AI's system prompt, which end users never see directly. It isn't published or indexed anywhere, so you can include internal pricing, qualification thresholds, or handling instructions you wouldn't put on a public page.
What language should the knowledge base be in?
English. If your clients interact mainly in another language, flag that early. It changes how the AI should respond to non-English messages and may need a different configuration.
What if my business has more than one distinct service area?
One knowledge base per AI agent. If you have genuinely separate service lines that shouldn't overlap in client conversations, separate agents, each with its own knowledge base, is the cleaner setup. Raise this early so the build gets structured correctly from the start.
Still stuck? Book a call.
And if the AI has already said something unfortunate to one of your better clients, email us before it drafts a proposal.