Skip to content

AI agents · Knowledge agent

Knowledge agent: find company knowledge instead of searching for it

A knowledge agent answers questions from your own documents — from SharePoint, file shares, email and the ERP. Every answer comes with a source. And everyone only sees what they are already allowed to see.

The situation

The everyday problem

"Where is the current inspection instruction?" "Have we dealt with this customer on this before?" "Who knows why we designed it that way back then?"

Questions like these cost every company time, every day. The knowledge exists, but it is scattered: in SharePoint, across three network drives with folder structures that grew over the years, in mailboxes, in the ERP, in the DMS. Search finds file names, not answers. So you ask the colleague who might know — and interrupt their own work.

On top of that comes something many managing directors in mid-sized companies worry about: experienced people will retire over the next few years. Much of what they know is written down nowhere.

How it works

What the knowledge agent actually does

  1. Connect the sources

    We connect the agent to the repositories that are authoritative in your company — usually two or three to start with, not everything at once.

  2. Inherit permissions

    The agent takes over the access rights from your systems. Anyone who cannot open a folder today gets no answer from it via the agent either.

  3. Understand the question

    Your people ask in plain language, with your terminology and abbreviations — not in search terms.

  4. Answers with sources

    The agent answers from your documents and cites the passage behind every statement. One click, and the original document is open.

  5. Honest about gaps

    If it finds nothing reliable, it says so — and shows which documents come closest.

  6. Make gaps visible

    Questions that have no answer are collected. That shows you where knowledge is missing and should be documented — for example before an experienced colleague retires.

Integration

Connected systems

SharePoint and Microsoft 365
Documents, Teams files, wikis
File shares
Folder structures that grew over time, including older file formats
DMS
Inspection instructions, drawings, procedures
Email
Shared mailboxes, for example for projects or service
ERP and CRM
Master data, order and customer history
Who decides

Approval stays with people

A knowledge agent reads and answers; it changes nothing. Which sources it may use and which it may not is your call. Personnel files, salary data or board papers stay out if you want them to — excluded technically, not just by instruction.

What the agent deliberately does not do

  • It does not bypass permissions. What an employee may not see, they do not see via the agent either.
  • It does not change or delete documents. It only reads.
  • It does not answer without a source. An answer that rests on no document is never presented as fact.
  • It does not assess people. It does not analyse who asks how much or who knows what.

On your own servers if you wish

By its nature a knowledge agent sees a lot of your company. That is why it can run entirely on your own servers: then no document and no question leaves your network. That makes it operable in line with data protection law even for companies with strict privacy and confidentiality requirements.

On-premise AI

Getting started

How to get started

01

First call, 30 minutes, free

We work out where most knowledge gets lost in your company.

02

Workshop

One day at your site. We look at which repositories are authoritative, how permissions are set up and which questions are asked most often.

03

The first agent

We start with a clearly defined area — quality or service, for example — and expand from there.

We run a knowledge agent in our own business. The building blocks exist — what differs in your company are the repositories, the terminology and the permissions.

FAQ

Questions about the knowledge agent

How is this different from SharePoint search?

Search finds files that contain a word. The knowledge agent answers the question — and shows which documents the answer comes from. It also searches repositories outside SharePoint.

How do you make sure nobody sees things they should not?

The agent inherits the existing access rights of your systems. On top of that, whole areas can be excluded. Who asked which question can be logged if your data protection officer requires it.

What about knowledge that only exists in people’s heads?

No agent can read that out. But it helps capture it: it shows which questions have no answer, and it prepares structured documents from conversations with experienced colleagues, which they then review.

Doesn’t AI sometimes make up answers?

Language models can, if you let them. That is why the knowledge agent is built to answer only from your documents and to back up every statement. If it finds nothing, it says so.

We already have an AI licence in house. Do we still need this?

Perhaps not. We will tell you honestly in the first call. A dedicated knowledge agent pays off mainly when your knowledge is spread across several systems, when your terminology is very specialised, or when the data must not leave the building.

How long does your team spend looking for answers?

In a 30-minute first call we work out whether a knowledge agent fits your company. Free.