AI knowledge base for business: how to set one up
Pick one group, list its most-asked questions and put the documents that answer them in one place, such as SharePoint. Add AI, and people get answers with sources. seef is a private AI assistant for your company's knowledge, in its own environment, hosted in the EU and run by a Dutch team, with the models agreed with you.
What is an AI knowledge base?
An AI knowledge base is a collection of company documents you can ask questions about in everyday language. You usually have those documents already: a folder of documents, a collection of work instructions or the files in SharePoint or Google Drive. The challenge is finding the right piece when you need it.
When you ask a question, the AI looks for relevant content in the connected documents. seef brings this together in a chat interface, with source references you can use to check the answer.
When is an AI knowledge base useful for your business?
If employees often have to search for agreements, instructions or earlier solutions, an AI knowledge base is worth exploring. The same applies when new colleagues need to learn how every folder is organised, or the same questions keep landing with the same experienced people.
seef fits organisations that want this without a technical team of their own. We set up and run the environment, so there is nothing for you to host or patch, and we help the people who will use it get started.
How do you create an internal knowledge base?
The same steps work with or without AI. Start small and expand based on use.
- Pick one goal and one group. For example, getting new starters up to speed, or answering the questions that always land with the same people.
- Write down the twenty questions that group asks most. That list decides what goes in first.
- Choose one place with a fixed structure, such as a SharePoint site, and decide who can read and who can edit. Copies kept in two places soon drift apart.
- Add the documents that answer those questions. Where an answer is not written down anywhere, write it up briefly.
- Give each topic an owner who keeps it current and removes old versions.
- Let the group test it for a few weeks. Check answers against the documents, note which questions go unanswered and fill the gaps.
Which documents should go in first?
Start with what people need every week: work instructions, procedures, templates, training material and project reports. Leave out drafts and outdated versions. With AI, they lead to confusing answers.
So you do not need to connect everything at once. seef connects to SharePoint and Google Drive, and you choose which files it may use. We usually start with a separate library that we set up and manage for you, holding only the files meant for AI.
Everyone with access to a knowledge base can get answers from the documents in it. So put in what is meant for those users. For confidential files, seef can set up a separate knowledge base with its own access, for example one per team.
seef is a service of Timformatie, certified to ISO 27001 and NEN 7510.
Read about securityHow do you use AI to search company documents?
You connect the AI to a source, such as a SharePoint library. The text of the documents is then indexed and updated when something changes. Employees can then ask questions about those company documents in the chat. For example: "What have we agreed about handing over a project?" The answer depends on what is in your own documents.
When an answer uses documents, seef shows source references. You click one, open the source document and read the text for yourself. That lets you check whether the answer fits the question and the situation.
A source reference is no guarantee that an answer is correct. Check important conclusions and keep the documents up to date. If an agreement has not been recorded anywhere, someone needs to add that knowledge first.
What can you do with seef beyond searching?
You can also use seef to summarise text and prepare recurring work. For example, you provide a meeting transcript and have it turned into minutes.
Want those minutes in the same format every time? You can create an agent: an assistant with fixed instructions for a particular task. You might specify that you want a summary, decisions and action items. You can download the result as a file, check it and share it.
Within your organisation, you can share agents and improve them together. That saves each employee from having to work out which instructions work well from scratch.
How do you keep track of AI use?
By deciding, as an organisation, which AI tools and which data employees may use. If employees choose their own AI tools, you want to know what information they share and which rules apply. Also agree which tasks AI should help with and which models fit those tasks.
One shared environment that everyone uses also gives you that overview: you know which tool is in use and which documents it can reach.
With seef, you agree in advance how you want the environment set up and which models you want to use. We also agree which data each service processes and where, up to processing only in the EU if you need it.
Read next: an AI policy template to copyHow does getting started with seef work?
We begin with a demo of the knowledge base, chat and agents, then discuss the setup, support and terms.
We often start with a group of five to ten employees. We help them prepare the first documents, try out questions and create agents for their work. This includes practical sessions where people can practise and ask questions.
A larger group can join after those first experiences. In our projects, this often happens after two to three weeks. We agree the pace with you. The first users can then help their colleagues get started.
What does an AI knowledge base from seef cost?
seef charges per user, with a minimum number of users. We discuss the specific price and minimum in the proposal. We also set out which setup work and support are included in the project.
Alongside the cost, allow time for your own team to choose documents, check answers and help colleagues.
What we see in practice
At a customer with no technical team, we saw people's hesitation about AI gradually ease. By practising together and sharing agents, even employees who initially had little confidence in it began to use it. That is what I find rewarding: knowledge that used to sit in folders becomes part of the work again.
Merijn van Es, co-founder of seef
Read how an adviser at Samergo works with seefSee what seef can do with your company's knowledge
In a demo, we show you how the knowledge base and agents work. Bring a few of the questions your team asks most, and we will discuss how seef would handle them and how you can get started.