The AI industry has spent years focused on the capabilities of models. Now, as enterprises move from experimentation toward deployment, another question is becoming harder to avoid: can those models actually reach the information businesses depend on every day?
Idan Shuster, co-founder and chief product officer of Unfold, believes that question represents a major piece of the enterprise AI puzzle. In a conversation with Omri Hurwitz at HumanX Amsterdam, Shuster described a market where organizations may have enormous amounts of proprietary information but still struggle to make it usable for AI because much of it sits inside complicated, legacy or proprietary systems.
The Data Exists. Access Is The Problem.
Unfold’s thesis is relatively straightforward. Enterprises do not necessarily need to generate more information before they can benefit from AI. Instead, they need to connect AI systems to information that already exists across their technology environments.
Shuster’s background gives him a particular perspective on that problem. He spent roughly a decade in cybersecurity, including hands-on penetration testing and offensive security work, before moving into product management at Varonis. He also spent four years in Israel’s Unit 8200 and later worked in cybersecurity services in Israel and the U.K.
That experience exposed him to both the technical and organizational sides of enterprise technology. “So I think both of them are kind of shaped the way into being a good product manager when understanding also like the business objectives, but also like the hands-on side of things,” Shuster said.
Unfold initially approached the problem through cybersecurity, helping teams obtain data for security and fraud-related initiatives. The company’s direction changed after a healthcare CISO introduced the team to a CIO who had a broader requirement: extracting information from proprietary systems to support AI workflows.
That conversation opened up a larger market.
From Security Tool To AI Infrastructure
Today, Unfold describes itself as an AI enablement platform. The company connects to enterprise systems and extracts information that can then be delivered into data lakes or directly into AI-agent workflows.
The challenge is that enterprise systems are rarely designed with today’s AI use cases in mind. A company might operate modern applications alongside ERPs, mainframes, healthcare platforms and highly customized internal software. Each system can contain critical information while exposing that information in very different ways.
Organizations have developed workarounds, but Shuster said those approaches can create their own problems. Enterprises may spend months requesting APIs from vendors or rely heavily on professional services to access information.
In some cases, the institutional knowledge is concentrated in a single employee.
“I have this one guy, and he’s the only one who’s familiar with the system. He works here for 30 years. The biggest fear is that he will live. What happens next?”
Unfold is designed to reduce that dependency.
Why Seven Days Matters
The company generally starts with one system rather than asking a customer to overhaul its entire environment. Shuster said the initial onboarding process, which Unfold calls “Unfolding,” typically takes around seven days.
The process is largely automated, but human verification remains part of the workflow because enterprise data cannot simply be treated as another software integration.
The company wants customers to see value immediately once the system is connected.
“With Unfold we are experiencing that once we are connecting a system, it’s like turning the light on. It was dark before, you never had the data. And now all of a sudden you can do whatever you want.”
That could mean querying the system, creating dashboards or feeding information into existing AI workflows.
The Business Case Goes Beyond AI
Shuster’s strongest examples are tied to measurable business outcomes. One healthcare customer acquires roughly 50 clinics each year, and integrating each clinic’s existing technology can take months. Removing that bottleneck could help the organization bring new acquisitions into its technology environment faster.
Another example involves a large retailer that relies on mainframes and needs to perform fraud analysis on information stored within them. In that situation, data access is tied not just to technology but to the company’s ability to expand.
Those use cases explain why Unfold is initially targeting large enterprises and Fortune 500 companies. These organizations tend to have complex environments, but they also have substantial incentives to make AI work.
Shuster expects the same challenge to spread as companies accumulate more software over time.
“I think that every company that has existing for at least five, 10 years and have both modern software and legacy software eventually will have this problem.”
The next phase of enterprise AI may therefore be less about adding another model and more about connecting the models companies already have to the systems they cannot afford to leave behind.
