Connecting SharePoint content to the AI knowledge base

SharePoint content can be connected to an AI knowledge base without the company changing systems: SharePoint remains the place for storage and collaboration, while the knowledge base connects defined libraries and adds an AI search on top. The team asks in everyday language and gets answers with sources – clicking the source opens the original document in SharePoint. No move, no duplicate maintenance, no break in the familiar way of working.

In brief

• In many companies SharePoint is established and will stay that way – connecting it to an AI knowledge base is not a change of system but an additional search layer.
• The built-in SharePoint search finds keywords and file names; the AI search answers questions in everyday language and backs up every answer with the source document.
• Selected libraries are connected; synchronisation keeps the content continuously up to date, and scanned PDFs are made accessible through text recognition.
• The system sets field values such as customer, project, installation or date itself; a person confirms the document type, because it determines who may see a document.
• Permissions are clarified before connecting, not afterwards: confidential areas stay out or are tightly restricted.
• The biggest gain comes when SharePoint is brought together with the other sources – mailboxes, file servers, voice notes – and one search covers them all.

Why is the SharePoint search not enough for many companies?

To be fair from the outset: as a storage and collaboration platform, SharePoint does a lot well. Documents are stored centrally instead of on local drives, several people work on files at the same time, versions are tracked, and via Microsoft 365 everything is linked to Outlook and Teams. Companies that have introduced it have taken a real step forward – and it would make no sense to tear that down again.

The disillusionment comes when searching. The built-in search finds file names and keywords – but if you ask “How did we set up the system at the Berger customer site?”, you get a list of hits instead of an answer and have to work through the documents yourself. If you use a different word from the document, you find nothing. And the content of old, scanned PDFs remains invisible, because technically a scan is an image. On top of that comes sprawl: after a few years, the knowledge is spread across team sites, subsites and libraries whose logic only their creators still understand. The knowledge is there – but it doesn’t answer.

What does “connect instead of migrate” mean?

Connecting means: the documents stay where they are. The knowledge base connects to the selected SharePoint libraries, reads the content and builds a searchable body of knowledge from it – comparable to a very thorough index for a library whose shelves nobody rearranges. Opening, editing and filing still happen in SharePoint; the knowledge base works with the content and points to the original with every answer.

This is the decisive difference from a migration, meaning moving the content into a new system. A move would be expensive and risky and would tear apart established processes – and it is unnecessary for the actual goal. After all, the goal is not new storage, but for the existing knowledge to finally answer questions. Nothing changes in the familiar way of working for employees; there is simply a new, much shorter route to information.

How does the connection work in practice?

The connection follows a manageable pattern:

1. Select libraries. Together we decide which SharePoint areas contain knowledge that should be available to the team via the search – and which stay out.
2. Set permissions. For each connected area, it is determined which roles in the knowledge base may see the content. Confidential material stays tightly restricted.
3. Make the content accessible. The knowledge base reads the documents. Scanned PDFs go through text recognition (OCR – converting an image into searchable text). The system takes over attributes such as customer, project, installation, date and topic by itself; what remains to be confirmed is the document type, because the document type determines who gets to see a document. This approval takes place in the review area and is a single click for unambiguous documents; the review area only asks questions where the document itself does not give clear information.
4. Keep it in sync continuously. New and changed documents are added automatically and go through the same process, and deleted ones disappear from the search. No second filing system is created that would need maintaining.

From then on, the same applies to SharePoint content as to all sources of the knowledge base: every AI answer names the source document, because AI can be wrong – the check is made against the original in SharePoint, one click away.

Who does what in future: SharePoint, ERP and knowledge base?

To avoid duplication, a clear division of labour between the systems helps:

System	Task	Typical content
SharePoint / Microsoft 365	Storage and collaboration	Current documents, team sites, joint editing
ERP / accounting	Leading system for documents and figures	Orders, invoices, delivery notes, payment status
AI knowledge base	Search and answers across everything	The how: instructions, reports, experience – with sources

The old rule of thumb applies here too: the ERP knows what happened – the knowledge base knows how it’s done. SharePoint, in turn, is the place where many of these how-to documents are created and stored. The knowledge base does not take away either system’s role; it makes their content queryable together – even across system boundaries, for example when documents for the same project are in SharePoint, emails in the mailbox and reports on the file server.

Where are the limits of the connection?

Honesty is part of it. First: connecting does not turn outdated content into current content. If SharePoint contains three versions of a work instruction, the AI does not know which one applies either – such contradictions become visible through the new findability and have to be decided once on the merits. Second: very old, self-hosted SharePoint installations from earlier generations are technically more complex to connect than Microsoft 365; this needs to be checked in advance. Third: with extremely poor old scans, even the best text recognition reaches its limits – heavily faded material stays faded.

And finally, data protection: connecting makes knowledge easier to find – including knowledge that was previously protected only by being impossible to find. That is why permissions belong at the start of the project. Goma-it’s AI knowledge base is designed for this with role management, data storage on EU servers and a data processing agreement (DPA); your documents are not used to train third-party AI models. The decision as to which areas are open to whom, however, remains the company’s – and it is made before connecting, not afterwards.

How do you get started with your SharePoint content?

The proven way in: connect one or two libraries of high everyday value – typically where instructions, reports or project documentation are stored – try them out for a few weeks with real questions from the team, then expand. This quickly shows which areas bring the greatest benefit and where permissions or old content need sharpening.

SharePoint is rarely the only source: only in combination with email mailboxes, file servers, scans and voice notes does the complete company knowledge emerge. Which sources there are and in which order connecting them has proven itself is shown in the overview Connecting documents and systems. How the introduction works as a project – from selecting the first library to everyday operation – is described on the implementation process page.