Unstructured data is defining the enterprise era of artificial intelligence. PDFs, videos, images and audio files now account for the vast majority of organizational content; yet companies have historically struggled to extract insights from this hidden trove. The rise of AI agents is turning that tide, making it possible to transform static files into dynamic sources of decision-making power.

The next wave of enterprise AI is being shaped by how well businesses can harness this vast pool of information to fuel more intelligent workflows and personalized recommendations, a shift increasingly powered by AI agents. That shift requires a new layer of technology — one capable of reading, classifying and reasoning over content the way humans do, according to Ben Kus (pictured), chief technology officer of Box Inc.

Box’s Ben Kus talks with theCUBE about how AI agents are evolving from basic bots into sophisticated collaborators.

“For Box, our whole lives, everything we’ve done for a long time all revolves around unstructured data,” he said. “In this world, thinking about not structured data and databases, but unstructured data, things like in files and content, images, audio, video documents. Our job is to secure it, to collaborate with it and then now to be able to provide AI on top of your unstructured data.”

Kus spoke with theCUBE’s Rebecca Knight and Jackie McGuire at Okta’s Oktane event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how AI agents are evolving from basic bots into sophisticated collaborators, why unstructured data remains the industry’s greatest untapped asset and how enterprises must build with governance and security at the forefront. (* Disclosure below.)

AI agents meet unstructured data

For most organizations, the ability to store unlimited files has led to a sprawl of unstructured content that’s difficult to manage. Classification has long been a costly and error-prone task, yet it is also essential for security and compliance. AI now enables the automation of classification at scale, tagging sensitive information and identifying high-value content across millions of documents, according to Kus.

“There’s so many places … where the more stuff you have, the harder it is for you to find and to organize and to do anything going forward,” he said. “At the same time, most customers … often don’t want to get rid of some of their old stuff.  For AI, it helps you in a couple of different ways. We’ve actually just announced our AI classification, where the AI can go through and basically be able to classify things and be like, from a security perspective or even just to be able to tag data that you say, ‘I’m interested in these kinds of things.’”

A second foundational step is retrieval-augmented generation, which enables AI to retrieve precise answers rather than just point to documents. By layering RAG on top of enterprise content, companies can quickly find case studies, customer histories or HR data that would otherwise require hours of manual search. This ability to deliver direct answers from vast repositories is becoming one of the most widely adopted enterprise use cases, Kus noted.

“What we see from many of our customers … AI could do so many things, and this is the beginning step,” he said. “They’re just taking a lot of their data, and they’re saying, ‘AI, can you find me this information?’ We see things like customers who use it for sales … which is if a customer has a question, you just ask AI, and it is able to look through tens of thousands of files, hundreds of gigabytes of data and then come up with not just, ‘Here’s a file to read it, but here’s the answer.’”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Okta’s Oktane event:

(* Disclosure: TheCUBE is a paid media partner for Okta’s Oktane event. Neither Okta Inc., the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE