A company knowledge base as a service: why it costs more than it seems
The client thinks the work is technical: upload the documents, connect it, done. The technical part is the smaller one. Below is what makes up the main part, why the contradictions found are sometimes valued more than the assistant itself, and how an internal version differs from an external one.
The client thinks the work is technical: upload the documents, connect it, done. The technical part is the smaller one. Below is what makes up the main part, why the contradictions found are sometimes valued more than the assistant itself, and how an internal version differs from an external one.
The same questions ten times a day
Any company of more than five people has a set of questions asked constantly.
How to process a return. What the terms are on this plan. What to write to a client asking to move a deadline. Where the current price list is. Who approves requests like this.
The answers exist. They sit in procedures, in message threads, and in the heads of two or three people who get interrupted all day because of it.
A new employee takes months to get up to speed for exactly this reason: not because the tasks are hard but because he does not know where anything is.
The solution looks obvious: an assistant that answers from the company's documents. But between "looks obvious" and "works" lies the part of the job people pay for.
What a knowledge base means here
Not a folder of files.
It is a connected source the assistant searches before answering. A person asks in his own words, the assistant finds the right place in your documents and answers from it.
The key difference from an ordinary chat: it does not retrieve the answer from the model's memory. It answers from what you put in. When all the material is supplied by you, there is nothing to invent — the most reliable mode of working there is.
Hence the other side: what is not in the base, the assistant does not know. And if it is not told what to do in that case, it will invent.
Why this costs more than it seems
The client usually thinks the work is technical: upload the documents, connect it, done.
The technical part is the smaller one. The main work is elsewhere.
Knowledge is scattered
The price list in one spreadsheet, a procedure in a document two years old, the current terms in the manager's messages, the exceptions in the head of a senior colleague.
Collection is not exporting files but an inventory: what exists, what is current, what contradicts what.
Half the knowledge is unwritten
The most labour-intensive part.
What people know has to be extracted by interview: what do you do, in what order, what goes wrong, how do you handle exceptions.
The person does not have to write anything, only answer questions. That is several times faster than a request to "please describe how you work", which usually gets answered in a month or never.
The documents were not written for search
A forty-page procedure in continuous prose searches badly. It has to be broken into meaningful chunks, each answering one question.
That is reworking rather than uploading. And it determines whether the assistant gives precise answers or general ones.
Contradictions
They turn up almost every time: the terms say one thing, the procedure another, the managers a third.
Finding the contradictions is a by-product clients sometimes value more than the assistant itself. It is what you bring them for nothing, and it is an excellent argument in the conversation about price.
How a knowledge base works technically
Briefly, so you can explain it to a client.
The first part is search across uploaded documents. The assistant receives a question, finds the appropriate chunks and answers from them.
File parsing — turning a file into text if the documents are not in text form.
Behaviour rules — what to do if the answer is not in the base. An obligatory point: say plainly that the data is insufficient and suggest asking a person. Without that rule the assistant will fill the gap with something plausible.
An updating rule — who refreshes the base and when.
Two different products
Worth separating, because the prices and the work differ.
An internal assistant for staff
Who uses it: your team.
What it solves: a new person gets up to speed in days rather than months. Experienced people stop being interrupted over trivia.
What is easier: the audience is friendly and an error is not critical — an employee will ask again.
An external assistant for customers
Who uses it: buyers.
What it solves: enquiries are handled around the clock, standard questions are closed with no human.
What is harder: the cost of an error is high. An assistant that gave a customer a wrong price or a non-existent condition creates a problem rather than solving one.
Two rules are obligatory here: do not invent and pass hot enquiries to a human. The sign of hot is a question about the price of an expensive item, about timescales, about individual terms.
It makes sense to start with the internal one. It is cheaper, safer, and the base gets debugged on it before going into the external one.
What is delivered to the client
A list worth showing in full — it also justifies the price:
- The inventory: what exists, what is out of date, what contradicts what
- The assembled and reworked base
- A list of the contradictions found, as a separate document
- A configured assistant with behaviour rules
- The rule for passing a question to a human
- Instructions for staff and training
- An updating procedure: who and when
The last two are skipped most often — and that is exactly why systems die within a month. An employee who has not understood routes around the system within a week. A base nobody updates starts lying within six months.
The order of prices
By my scrape of 5 986 jobs over four months: a system or a pipeline $490–1 460, against a median one-off job of around $60.
A caveat: an explicit budget appears in 17% of listings, and by individual type of work there are between twenty and three hundred observations with a sum.
What determines the price: the volume of documents, how much knowledge has to be extracted by interview, one department or several, an internal assistant or an external one, whether maintenance is needed.
What to compare it against in conversation. Not the price of software but time: how many hours a month your experienced staff spend answering colleagues, and what a month of getting a new person up to speed costs.
Where the main money is: in maintenance. Terms change, new questions appear, the base goes out of date. That is a recurring payment and an honest one.
What the platform offers for a knowledge base
For regular updating, use scheduled launches in n8n and Make chains: ready templates are in the automation templates section (Creator and above).
200+ AI assistants across 14 categories — for interviews and reworking documents. HR will help extract knowledge from people, legal will check the wording. From the Basic plan.
The Bot architect from the "Engineers" category — a system instruction with behaviour rules, handover triggers and a constraints block.
Creating your own assistants with a separate base — the Creator+ plan.
Automation templates — if a link with the company's existing systems is needed. Creator plan, advanced ones Full.
Registration is free and opens three days of full Basic access.
Where to start today
One action. Take a company you know — your own or a client's — and write out the five questions employees ask each other most often.
Then find where the answers are written down. Usually it turns out two of the five are written nowhere, and another two are written in two places differently.
Those four points are your first sale. Not "let us build a knowledge base" but "you have four questions the company has no single answer to".
What RAG is and why people pay for it
The "Automations in n8n" course (N8N plan) devotes a module, "RAG agents", to this, and the phrasing there is direct.
A RAG agent is a model trained on a company's private base: PDFs, sites, procedures, correspondence. It answers from internal documents rather than from the internet — with no invention.
The phrasing from the course: you create an AI that knows the company's entire documentation by heart.
Why that is expensive. Large businesses want their own chat assistant that does not take data outside and does not make things up.
The next level: agentic RAG
Ordinary RAG retrieves information and answers. It has one source, one attempt and no recheck.
Agentic RAG works differently: it dynamically selects several sources, refines the query if the first result is weak, rechecks what it found and makes a second pass.
What that changes for the client. An ordinary knowledge base answers direct questions. This one handles complex ones where the answer is assembled from several documents.
Ready-made RAG templates on the platform
Adaptive RAG: answers from a knowledge base (n8n): classifies the question and picks one of four search strategies over the corporate base — the closest thing in the catalogue to a "smart" RAG. An ordinary template: Creator, Make, N8N, Full. Other chains — RAG chatbot with a Google Drive knowledge base and RAG search over a Notion knowledge base with Supabase sync.
Where this is assembled with no development
n8n and Make automation templates, Creator and above — ready chains for document tasks: import, plug in your keys, run.
The "Automations in n8n" course, N8N plan — the deepest level: your own infrastructure, your own bases, no subscription limits and no data going outside.
The Open Source section — if the data must not leave the client's perimeter at all: over 800 projects, including knowledge bases and corporate search. From the section's description: critical for law, medicine and finance.
Why this costs more than a bot, in one sentence
A bot answers from one source. A knowledge base has to work out that the procedure from March contradicts the email from June and say which of them applies.
Assistants for gathering material, Basic plan: Academic editor — bringing assorted documents to a single form. Learning and development — if the base is for onboarding. Customer Success expert — if customers will be answered from it.
Transcribing what was never written down. The Gen AI section, "File → text" mode: half a company's knowledge lives in meeting recordings rather than in documents.
Skills for assembling a base
The article explains why this costs more than a bot. Below is what that more expensive work consists of. Almost all of these skills are advanced: Creator+, or $9 one-off, or 1,000 XP; only Document360 is available from Basic.
- Designing systems with vector search — the whole pipeline step by step: chunking documents, embeddings, choosing storage, hybrid search and reranking. And an honest answer to the question the client will ask first: why does the search find the wrong thing
- Semantic search with an index — the technical part: searching by meaning instead of "put all the text in the request"
- Synthesising knowledge from different sources — when the same thing sits in a chat, an inbox and a document: deduplication and choosing the freshest version. At low confidence it honestly warns that the data may be out of date
- A local knowledge base — a variant with no vector storage and no uploading anywhere: a folder, navigation, rules for the agent. A check for broken paths and version conflicts
- A knowledge graph from documents — it takes various formats and builds links with a confidence score. Useful when the client does not know what is in his own archive
- Base analysis and graph building — it finds hidden dependencies between documents
- Notion — knowledge base management — if the client's base is already there
- Document360 — if what is needed is a self-service portal for customers rather than internal search
What to take on the platform for this task
- Adaptive RAG: answers from a knowledge base — n8n template: question classification and choice of search strategy; Creator, Make, N8N, Full.
- Telegram knowledge-base bot with RAG — n8n template: accepts files, splits them into fragments and answers from them; Creator, Make, N8N, Full.
- MaxKB — a corporate AI agent on your own data — Open Source: RAG, workflows and MCP tools on your own servers; description open to all, repository link from Creator.
- DocsGPT — an AI assistant for your documents — Open Source; repository link from Creator.
- Documize — a corporate knowledge base on your own server — Open Source; repository link from Creator.
- Automations in n8n — course, "RAG agents" module; N8N plan.
What to read next
What an assistant for a department costs — how to price work like this.
The anatomy of a system instruction — the technical part of the project.
Auditing procedures — what is done before gathering knowledge.
Local models for sensitive data — if nothing may leave the building.
n8n and Make automation templates — the full guide to automations.