A blog by Bastiaan Brefeld
For a long time, a DMS system was assessed on the basis of roughly three questions. Can you save a document? Can you retrieve it? Do you know which version is current? Anyone who could answer ‘yes’ to these questions had their document management in order. The bar has now been raised. Your DMS now also determines who is authorised to view what, whether you can demonstrate during an audit what has happened to a document, and whether the AI tools your organisation applies to it can extract any reliable information.
The latter is, of course, very much in evidence at the moment. Gartner surveyed between May and July 2026 more than 2,000 CIOs and IT leaders. Only 13% said that their organisation had derived significant value from AI to date. At the same time, BCG in September 2026 It is clear that a small group of frontrunners (7.5%) are pulling ahead and generating 2.4 times as much revenue growth as the ‘laggards’. And this is despite the fact that everyone has roughly the same tools at their disposal.
So what are these frontrunners doing differently? They didn’t start with the AI tool. They started by asking where that tool gets its information from and how reliable it is.
The same canal-side house, a different foundation
I’ll use the example of two canal-side houses in Amsterdam. Both have been beautifully restored. One will still be standing in a hundred years’ time, whilst the other is slowly leaning to one side because the wooden piles underneath are starting to rot. You can’t tell the difference just by looking at the façade. It’s the same canal-side house, but with a different foundation.
It’s the same with AI. The tools are the same everywhere, but it’s the foundation that determines whether they’ll deliver results. And most of an organisation’s knowledge is contained in documents. A process often begins with a document: a contract, an email, an agreement. It also ends with one: a report, a recommendation, an invoice, a file. You name it. In between, there are meeting minutes, memos, drawings, analyses and work instructions.
Gartner analyst Mark Beyer said in May 2026 To put it simply: 70 to 90% of the data within an organisation is unstructured, such as documents. And research firm BARC found in early 2026 that, according to 70%, less than half of that unstructured data held by organisations is discoverable and usable for AI.
Why is simply storing documents no longer enough?
For years, a DMS system revolved around storage, retrieval and version control. That was enough as long as people filled in the gaps. A colleague knows which contract was signed most recently and which client a report relates to. A document does not know that. It cannot explain itself.
AI can’t do that either. If you let an AI tool loose on folders full of files without any context, you’ll get summaries and faster search results. That’s handy, but it won’t pay for your Copilot licence, nor will it allow you to get the most out of your Microsoft 365 investment. It only becomes interesting when AI can, for example, tell you which version is current, who approved it and what’s still outstanding in a file.
To achieve this, your DMS needs to do more than just store documents. It needs to provide context: each document must be linked to the customer, project or process to which it belongs. We call this metadata. It is the difference between a digital filing cabinet and a foundation.
What DMS questions should you ask yourself?
Not: how many documents can we store? But rather: does our system know which client or project a document belongs to, which version is current, and who is authorised to view it? If the answer is no, then in my opinion no AI tool will be able to sort that out for you. AI merely makes a weak foundation more apparent more quickly.
On 8 December, I’ll be giving a 20-minute webinar about the role of your DMS in the world of AI, security and compliance. I offer practical advice, whether you’re already using SharePoint, are considering M-Files, or are still exploring your options. Sign up here in.




