AI knowledge management: every answer with its source.

Knowledge management is capturing, organising and making findable the knowledge in an organisation, so everyone works with the same, up-to-date information. That knowledge sits in documents, systems and the heads of experienced staff. AI knowledge management makes it searchable with ordinary questions: an AI model first finds the relevant passages in your own sources and then builds an answer on them, with a reference to the document.

The documents are split up and indexed by meaning, not just by keywords. When you ask a question, the system retrieves the passages that deal with it and the language model turns them into an answer. That technique is called RAG (retrieval-augmented generation). Because the answer comes from your own sources, you can always check it.

Document management is about storing, organising and securing documents: folders, versions, permissions and retention periods, for example in SharePoint or a case management system. Knowledge management goes a step further: making the knowledge in those documents findable and usable. An AI knowledge base builds on your document management. It uses the files, versions and permissions that already exist, and gives an answer instead of a list of files.

Usually not. Traditional knowledge base software needs someone to write the knowledge up as articles and keep them current. An AI knowledge base works on the documents you already have, for example in SharePoint, Google Drive or your case management system, and picks up new versions automatically. Where knowledge has not been captured anywhere yet, we start by capturing it.

No. General-purpose assistants know a lot about the world, but little about your internal policies, procedures or client files, unless those have been properly made accessible. A knowledge base on your own sources gives answers that are right for your business, with source and version. It can sit perfectly well alongside Copilot or ChatGPT.

Anything that holds knowledge: policies, handbooks and procedures, Word and PDF files, SharePoint and Google Drive, wikis, tickets and email, and data from your systems such as your CRM or case management. We start with the sources that most questions are about.

By letting the model answer only on the basis of the passages it has found, and always showing the source. If the system cannot find a good source, it says so instead of guessing. In the early phase we test with real questions from your staff and refine until the answers are right.

Yes, if you connect it to the place where documents are updated. New versions are picked up automatically and old versions are no longer used. At HypotheekCompany every answer shows the source excerpt, the lender and the document version, so staff can always check what the answer is based on.

We set up the knowledge base so it follows the permissions you already have: anyone who may not open a document gets no answer from it either. We work with business AI models that do not use your data for training, choose processing within the EU wherever possible and record that in a data processing agreement.

By capturing knowledge at the moment it is created: in conversations, case files and decisions. At In het Veld, conversations are automatically stored, transcribed and turned into action points, with one central source of truth for the whole team. Whatever is captured that way can then become searchable in a knowledge base.

That depends on the number of sources, the volume of documents, the integrations with your systems and the number of users. We agree a fixed scope and a fixed price up front, and a first version on a defined set of sources is often live within a few weeks. Keeping it up to date runs on a predictable monthly fee. Book an intro call for an estimate for your situation.

A knowledge base is often the first layer of an AI Operating System: first make the knowledge findable, then connect systems and let AI take over work. You can start with the knowledge base without building the whole system straight away, and extend it later where it adds value.