Automation · People, data and knowledge
Second brain: a company knowledge base for AI
We build a second brain – a company knowledge base made from your procedures, price lists and policies – and connect it to the AI you already use. People ask in their own language and get an answer with a link to the document it is based on.
Sound familiar?
According to the Czech Statistical Office (ČSÚ), in 2026 half of medium-sized and just under a third of small companies used artificial intelligence. But AI that does not know your procedures, price lists and agreements gives generic answers.
-
“Only a few people know everything, and when they leave, we have a problem.”
-
“A new colleague keeps asking the same questions for three weeks.”
-
“We have five versions of the price list on the drive and nobody knows which one is valid.”
-
“We paid for Copilot and nobody uses it.”
-
“People put company data into ChatGPT and we do not know what to do about it.”
-
“Machine procedures are on paper next to the machine, or nowhere.”
Flow
How the assistant answers – and where we keep watch
The assistant answers from documents the person asking is allowed to see. Your people decide what in them is valid.
- 01
Question
An employee asks in their own language, in a tool they know.
Person: They do not need to know which folder the document is in.
- 02
Permissions
The assistant sees only what the person asking sees.
Person: You decide what does not belong in the knowledge base at all, such as payroll or personnel files.
- 03
Search
It searches approved documents that have an owner and a review date.
Person: The owner keeps their page up to date.
- 04
Answer
A short answer with a link to the document it is based on.
Person: For important matters, they open the source and check it.
- 05
Gap
When the answer is not in the materials, the assistant says so and the question goes to a list of missing guides.
Person: The owner adds the guide.
What we build for you
An AI assistant for businesses is only as good as the materials it answers from. That is why we start with knowledge, not with a chatbot.
Company wiki in your tool
A structure on Google Drive, in SharePoint, Notion or Obsidian. Templates for procedures, decisions, onboarding a new hire and frequently asked questions. Every page has an owner and a review date.
Writing standard operating procedures (SOPs)
We write up procedures from interviews and recordings, and each one is approved by its owner. We convert paper instructions and merge duplicates.
Digital work instructions at the machine
A QR code at the machine or workstation opens the current version of the procedure, with photos, on a phone. Who owns the procedure and when they last reviewed it is visible right away.
AI assistant over your documents
It answers from verified company sources and shows the document each answer is based on. When the answer is not in the materials, it is set up to say it does not know. Built on Gemini, Copilot, Claude or as a custom solution.
Connection to company systems
Context from Raynet, POHODA or ABRA Flexi, for example “what terms does client X have”. Always read-only and with limited permissions.
AI rules, policies and training
How the AI should write, what it may do and when it should ask a person. Plus an internal policy on which tools people may use and what not to put into them, and a short training for the team.
Maintenance and additions
The knowledge base grows with the company. We watch for outdated pages and send an overview of questions the assistant could not answer. That becomes the list of missing guides.
What we build it on
First, what you already pay for. Gemini in Google Workspace for businesses, Microsoft 365 Copilot and Claude Team can all work with company documents. What matters is what is in them and who has access to them.
Google Workspace and Gemini
Gemini in Gmail is in all plans; in Docs, Drive and Meet from the Business Standard plan. Gemini Notebook answers from the sources you add to it. Google states that it does not use company data to train models.
Microsoft 365 Copilot
It answers across Outlook, Teams, SharePoint and OneDrive, from data the user has access to.
Claude Team and Notion Business
Claude Team has connectors, enterprise search and Claude for Microsoft 365. Anthropic states that by default it does not train on your content. Notion Business has Notion Agent and search that also covers connected apps.
Obsidian
Notes are plain Markdown files in a folder the company owns. Paying for a commercial license is not required.
Custom assistant (RAG)
A custom assistant over Google Drive, SharePoint, Notion or a folder of Markdown files. It keeps up with changes to your documents. Runs in an EU cloud or on your own server.
Raynet, POHODA, ABRA Flexi
Raynet has an official MCP server for AI assistants (MCP is a standard through which AI reads data from other systems). Open-source MCP servers exist for POHODA and ABRA Flexi. The assistant can then also answer from data in the CRM or accounting.
Are your documents scattered across emails and binders? That is a starting point too. We do not add personal mailboxes to the knowledge base across the board, only selected shared mailboxes.
When you do not need us
We do not sell licenses. When the tool you already have is enough, we tell you.
- You pay for Google Workspace Business Standard or Notion Business and your documents are in order
- You already have Gemini, or Notion Agent, in your plan. Try what you pay for first. Get in touch when the AI starts answering from old versions.
- You need answers from a few documents for one team
- Gemini Notebook over selected sources is enough for that. It is included even in the basic Google Workspace plan. You do not need to build a knowledge base for it.
- You are looking for a customer chatbot for your website
- That is a different service. This is about knowledge inside the company: procedures, price lists, policies and answers for your people.
Second brain at our company and at clients
PackLabs second brain
We run our own company, PackLabs, on a second brain that AI works with. In it, the AI finds how the company works, who we work for and which rules it must follow, and uses that to prepare orders, plan and sort documents. We build the same approach for you.
Company knowledge and work structure
For FS Composites we built the company’s knowledge and work structure and automated routine tasks.
Automation referencesHow it works
- 01
Free audit
30 minutes. How you use AI today, where your know-how is and what slows you down.
Output: Recommended next step
- 02
Knowledge audit
We go through drives, documents and sharing. With the people who know the company, we also write down what exists only in their heads.
Output: Knowledge map and proposed structure
- 03
Pilot
One team, one area. We build the structure, write down the procedures and test the assistant on your real questions.
Output: Verified questions and answers
- 04
Rollout
Further departments, permission setup and a short training for the team.
Output: Knowledge base and assistant in operation
- 05
Maintenance
Checking outdated pages and an overview of questions the assistant could not answer.
Output: Ongoing support (optional)
Frequently asked questions
What is a second brain, or a knowledge base for AI?
Do we need new software for this?
Won’t a colleague see payroll or contracts through the AI?
Where is the data and is the AI trained on it?
What if the AI gives a wrong answer?
Is it a RAG chatbot?
Is Obsidian suitable for businesses?
What does the AI Act require of us?
How much does it cost and how long does it take?
Related
Moving to Google Workspace
Company accounts, shared drives and email on your domain. The foundation Gemini then runs on.
MoreRecruiting and onboarding
A new hire starts without explaining the same things over and over.
MoreDocuments and contracts
Templates, generation and a register of contracts.
MoreData migration and cleansing
Data without duplicates that answers can be based on.
MoreEnquiry — AI knowledge base
Tell us what you are dealing with. We start with a free 30-minute audit.