Panzura LLC, maker of a global cloud file system, today announced the general availability of Nexus, a new platform designed to connect enterprise file data to Microsoft 365 Copilot while preserving existing security controls and permissions.

The launch addresses what the company said is a longstanding limitation in enterprise artificial intelligence: Most unstructured data stored in file systems has been largely inaccessible to large language models.

“Customers have tons of data that they historically have not been able to leverage because it’s by nature unstructured,” said Chief Executive Officer Karthik Ramamurthy.

The Nexus platform integrates with Panzura’s CloudFS hybrid cloud file system to expose data through Copilot’s conversational interface. The goal is to allow users to query enterprise knowledge using natural language without changing existing workflows or data architectures.

“Copilot cannot see the bulk of the enterprise data,” said Mike Harvey, senior vice president of product. “Nexus changes that.”

Unlike rival approaches that rely on periodic indexing, Nexus uses an event-driven architecture to synchronize data and permissions in near-real time. The system captures file changes and access control updates as they occur, reducing the lag time between updates that can create inconsistencies between source systems and AI outputs.

Policy preservation

That’s a security problem, Harvey said. “Those changes need to propagate through the system in near-real time.”

The platform ingests selected data based on administrator-defined policies and maps file system permissions directly into the Copilot environment. Panzura said this ensures that users can access only the information they are authorized to see when interacting via AI.

“If you don’t have access to finance data, you can ask questions, but you won’t receive any of the knowledge,” Harvey said.

The architecture avoids full data scans by relying on audit event streams already generated by enterprise file systems. These events are used to update an underlying knowledge graph that supports Copilot queries.

“We capture every single event, and map that to the policies,” Harvey said. “We make those changes to the graph structure on the Copilot side.”

Data is broken into smaller units, or “chunks,” and embedded with associated security metadata before being indexed. “Any chunks that you don’t have access to don’t even become part of the conversation,” Harvey said.

RAG at scale

Panzura said the result is a system that can support retrieval-augmented generation at enterprise scale while maintaining strict governance controls. Early use cases have emerged in industries such as architecture, engineering and construction, where organizations manage large volumes of project documentation. Customers use the system for tasks such as proposal generation, supplier risk analysis and workforce planning, which require accessing and comparing large number of documents.

Harvey said an example query might be, “Find me all of the proposals that were successful and give me the content that made them successful.”

The platform also enables more complex analysis across multiple documents, allowing organizations to identify patterns and insights that were previously difficult to surface. “When you give the customer the ability to talk to the knowledge of the organization, it becomes active knowledge,” Harvey said.

For now, Nexus is tightly integrated with the Microsoft ecosystem, reflecting widespread adoption of Copilot among enterprise customers. Microsoft has said more than 70% of Fortune 500 customers have adopted Copilots, although usage rates remain low.

Broader support is planned. “More options will exist in the future,” Harvey said.

Panzura positioned the launch as an initial step toward enabling more advanced AI workflows, including agent-based automation built on top of enterprise data.

“This is the first of many innovations to come,” Ramamurthy said.

Image: SiliconANGLE/Google Whisk