Free tools Windows power users keep installed
One-click scans. No signup required.
Enterprise AI can be limited not only by what a model can do, but by whether it can reach and correctly interpret information inside a company’s existing systems. That was the central argument in a conversation between Omri Hurwitz and Unfold co-founder and chief product officer Idan Shuster at HumanX Amsterdam, as reported by TechBullion on October 2, 2026.
Why access to enterprise data is an AI problem
Shuster’s argument is that many companies already hold valuable information, but it may sit in proprietary, legacy or otherwise difficult-to-access systems. Those systems can expose data inconsistently or lack convenient APIs, making it hard to bring information into AI workflows without first solving an integration problem.
As an Amazon Associate I earn from qualifying purchases.
This is different from a model simply needing more data. The question is whether an AI system can reach information that a business depends on, and whether that information arrives in a form the system can use. Interview coverage says Unfold initially pursued data access for security and fraud-related work before broadening its focus after a healthcare CISO connected the team with a CIO seeking data from proprietary healthcare systems for AI workflows. Those company-history details are reported by TechBullion; they are not independently verified here.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Why raw records are not the same as context
Even when an agent can reach a database, a collection of records may not explain how people use the application or what its data means in a particular workflow. Interfaces, custom components, business processes and underlying system structures can all affect how records should be interpreted.
#1 Best Overall
As Shuster put it, “If you let the agents interact directly with the data, sometimes it doesn’t make sense,” in a quote reported by The San Francisco Tribune on October 1, 2026. The point is that access alone may not preserve the context users get from working in the application itself. Connecting an agent to data without accounting for that context risks producing answers that are technically based on available records but not useful for the task at hand.
How Unfold says its integration layer works
Unfold describes itself as an integration layer for enterprise systems, including systems without APIs or exports. Its current product page says a customer points the service at a system, Unfold understands it layer by layer, and then provides normalized, governed outputs for existing tools and AI workflows. The company names Splunk, Cortex, Microsoft OneLake, Snowflake, Databricks and AI agents among the systems and destinations in that surrounding stack.
This is Unfold’s product positioning, not independent evidence that it supports every system or performs equally well in every environment. The company says initial work on one system typically takes around seven days, and its site uses the phrase “Any system. Live in 7 days.” Treat that as a company-reported timeline for initial system onboarding, not a guaranteed service level or independently measured benchmark. TechBullion’s account also describes human verification during onboarding.
Examples Shuster used to explain the problem
Bringing acquired clinics into a healthcare organization
In an interview example reported by TechBullion and The San Francisco Tribune, Shuster described a healthcare organization acquiring roughly 50 clinics a year while integration of each clinic’s existing technology takes months. The example illustrates how differences among established systems can slow the process of making information usable across an organization. The publications do not name the customer or provide independently audited results, so the example should not be read as proof of specific savings or faster integration.
Using mainframe data for retail fraud analysis
Shuster also described a large retailer that needs mainframe data for fraud analysis. The case illustrates the value a business may see in information held in an older system, even when it is awkward to connect to newer AI workflows. The interview coverage does not identify the retailer or quantify the impact of the work.
What the conversation establishes—and what it does not
The interview presents a case for treating data access and application context as practical constraints on enterprise AI adoption. It also offers examples of the kinds of systems and workflows that can create integration challenges. It does not provide an independent technical evaluation of Unfold, audited customer outcomes, or a comparison with other products. The examples support understanding the problem Shuster is addressing, not a product ranking.
The reporting says Unfold initially targets large enterprises. That focus fits the interview’s emphasis on organizations operating a mix of modern and legacy software. Shuster’s broader claim, quoted by TechBullion, is that companies with both types of software may eventually encounter the problem; it is his view, not an established forecast for every business.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Questions to ask when evaluating this kind of approach
The conversation does not compare vendors, but its central problem suggests practical questions for an enterprise team assessing any integration layer for AI:
Best Value
- System coverage: Can it connect to the specific proprietary or legacy systems the organization needs, including those without convenient APIs or exports?
- Context mapping: Does the approach account for workflows and application behavior, or does it only extract underlying records?
- Governance and review: What controls govern normalized outputs, and where does human verification occur?
- Destinations: Can the resulting data reach the tools, platforms and AI workflows the organization already uses?
- Evidence of outcomes: Can the vendor provide customer results that are independently verifiable and relevant to the systems and use case under consideration?
These are evaluation questions derived from the integration challenge described in the interview, not criteria tested or scored by its reporting. Shuster’s background in cybersecurity, including penetration testing and offensive security, and subsequent product-management work at Varonis are also reported in the interview coverage; TechBullion additionally reports prior service in Israel’s Unit 8200. These biographical claims have not been independently checked here.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
