From 100 million files to real insights

The financial sector has been adopting technology decades earlier than other industries. Banks have long since ceased to be just banks; they are IT organizations that happen to offer financial products as well. Customers expect to file a claim in a minute via an app, receive a mortgage offer via video, and get immediate answers without any wait time. That’s no longer a luxury—it’s the standard.

Organizations that embraced technology early on now have a structural competitive advantage. The question is: How can they leverage that advantage in a smart way?

Thirty years ago, content consisted of scanned documents and PDF files. Today, content can be a screenshot of a Facebook page, a voicemail message from a customer, a personalized mortgage video, or a handwritten form. The form in which information appears has changed radically, but the way many organizations handle it has not.

That’s where the problem lies. Because you have to store, classify, and secure all that information—and get it to the right person at the right time. If you don’t do that in a structured way, you don’t have information management—you have a digital junk room.

Roberto puts it bluntly: just because you have 100 million files stored somewhere doesn’t mean you know what’s in them. And if you don’t know that, you can’t do anything with them. No better processes, no higher customer satisfaction, no better decisions.

That’s what Roberto calls “dark data”: information that exists but remains hidden. OCR has long been used to make documents searchable. But searchable doesn’t mean understandable. AI is changing that—not by archiving documents, but by analyzing them, summarizing them, and linking them to the knowledge already stored in your systems.

Imagine this: a lawyer reviewing a complex claim no longer has to sift through hundreds of pages before forming an initial opinion. AI reviews those documents, creates a summary, and provides references to the sources. The lawyer makes the decision; the AI does the groundwork. That’s not science fiction—it’s what OneXillium’s clients are already doing today.

What makes this possible is the structure that was put in place years ago. Customers who have properly organized their digital archives from the start with metadata, tags, and a consistent structure can now apply their own Large Language Model to that data. They ask questions of their own document library and receive useful, traceable answers. It’s like having your own ChatGPT, but one that’s fed exclusively with your own, reliable content.

Roberto introduces a distinction that sticks: the difference between a knowledge worker and an intelligence worker. A knowledge worker possesses expertise. An intelligence worker combines that expertise with AI support and, as a result, works more efficiently, faster, and more accurately on a consistent basis.

What stands out: employees with less experience often benefit the most from this. They tap into the knowledge embedded in the system and are thus able to deliver high-quality work more quickly. This represents a strategic opportunity for any organization facing a tight labor market and an aging workforce.

Anyone who thinks that AI implementation starts with choosing a tool is mistaken. Roberto is clear: start with policy. What do you want to achieve? Who has access to what information? How do you protect sensitive data? How long do you keep what?

The financial sector has an advantage in this regard, even though it may not always seem that way. Laws and regulations such as DORA, AFM guidelines, and those of the Dutch Central Bank compel organizations to implement robust information security measures. That framework may seem like a hindrance, but in reality, it serves as a foundation. Those who lay that foundation properly can then build on it freely. Other sectors do not (yet) face that pressure, but that does not mean they can afford to wait. Frameworks such as ISO 27001 offer a solid starting point even without a legal obligation.

Roberto concludes with a message that leaves little room for debate: a 5 to 10% increase in productivity is realistic for organizations that use AI in their information management. The investment required to achieve this doesn’t have to be enormous. But taking that step is essential.

Here’s what you can do—starting tomorrow: identify the employees in your organization who have a knack for technology. Give them the freedom and time to experiment. Ask them what they would improve. After all, your people know better than anyone where the bottlenecks are in your processes. As a director, all you have to do is create the right environment.

If you wait until the technology is mature enough, you'll wait too long. What wasn't working well just six months ago is already working exceptionally well. And in six months, the same will be true of what still seems experimental today.

Listen to the full podcast and learn more about the specific projects Roberto describes.