Commvault Tackles AI Data Requirements With New Data Room, MCP Server

‘This is less about how we are using AI in product delivery and others to kind of make life simpler and easier for users. … As we’ve been adding more awareness and context and other things into the data sets, how do we also start to expose them in an easier way to interconnect with AI projects,’ says Brian Brockway, Commvault’s global CTO.

Data protection and cyber resilience technology developer Commvault has unveiled a new secure environment to connect trusted backup data to AI platforms or other AI initiatives along with a new policy-based bridge between enterprise systems and GenAI systems using Model Context Protocol (MCP).

The new technologies are focused on data resiliency for the growing amount of data being used with AI, said Brian Brockway, global CTO for Tinton Falls, N.J.-based Commvault.

“This is less about how we are using AI in product delivery and others to kind of make life simpler and easier for users. … As we’ve been adding more awareness and context and other things into the data sets, how do we also start to expose them in an easier way to interconnect with AI projects,” Brockway told CRN.

Commvault’s new Data Rooms form a secure environment where enterprises can keep trusted backup data and make it available for use with their AI platforms or internal data lakes, Brockway said.

“With a lot of the big AI projects today, more and more customers are looking to put their own spin on the data, whether it’s vector databases or RAG or other elements, to get rid of hallucinations or do a little bit more curation of the data,” he said. “That’s where you have to say, ‘How do I couple some of my data assets to be able to better intersect and connect them forward with some of my AI projects?”

Commvault for a number of years has had the capability to add more intelligence and semantics into the environment by indexing it and understanding it for more vectorized and multivariate capabilities and move from a Google-style search to getting a view of content and data sets, said Brockway (pictured above).

“That’s the data room, a virtualized data sharing relationship,” he said. You can plug into the data and use it either for your own internal AI applications or work with third parties where you need to add some PII (personally identifiable information) sanitization and data masking or other elements.”

This is not the same as setting up a file sharing system, Brockway said.

“There’s the element in the middle of saying, how do I bring a collection of data forward and then be able to present it to the third-party AI system,” he said. “But in a lot of cases, corporate data can have a lot of heritage of where it’s all come from. There’s a variety of different security models that may be intersected. And sometimes when we’re working inside the four walls of the enterprise, we maybe don’t pay close attention to what’s in the data, such as PII. … We’re saying, how do we take that data set and rather than just clean it and provide access, how do we virtualize the access point in the middle.”

Commvault’s Data Room comes as IT groups have had access to data for years but have struggled to crack it open to create value, Brockway said.

“How do we crack this open without having to do a tremendous amount of movement and migration,” he said. “The last thing anybody wants to do is to create four different repositories in the same data. They’d like to say they have one managed collection at the center and then expose access and rights, maybe on a short, temporary basis, maybe with some masking in between, and evolve that data set to make it more friendly to future AI applications.”

Commvault’s new Model Context Protocol server serves as a bridge between enterprise systems and common GenAI systems to let users interact with the Commvault Cloud using simple natural language.

MCP is an open-source standard for connecting AI applications to external systems to allow AI applications like ChatGPT or Claude to connect to data sources, tools, and workflows to access information and perform tasks.

A standard like MCP is important to ensure having a conversation with AI doesn’t result in things like accidentally ordering a pizza, Brockway said.

“When we started using more generative AI in our environments and user experience, that’s when we started to really break down, I’ll say, the experience barrier,” he said. “So rather than you have to have a Commvault black belt to onboard something or know the prescriptive way to do it, we can be a little bit more conversational by nature. Tell us what you’re trying to get done, and can we guide you through that experience?”

Brockway said to take a web-based application like Docusign, for instance, where after negotiations are done, the documents are signed, and then you get the option to save the documents. They have to be captured and placed into retention repositories and made available for search, which may require the IT team to help.

The Commvault MCP server connects into the Commvault environment through MCP so Claude or ChatGPT would talk to the Commvault AI engine to immediately onboard and store some of these new content pieces just created on the Docusign service with the specific policies applied, he said.

It is hard not to do more and more with AI, Jay Waggoner, president of Edge Solutions, an Alpharetta, Ga.-based solution provider and Commvault channel partner.

“Customer demand is driving all of us to kind of figure it out,” Waggoner told CRN. “I do think we’re still in the early days. I think as an industry we’re still all figuring it out. And I think the wave is still very much in front of us. We’re definitely figuring it out ourselves and helping customers to figure it out along the way as well.”

Waggoner said his company has a long history with Commvault.

“To their credit, they’ve really been in a state of reinvention over the last several years, and they’ve done so very successfully,” he said. “Others entered the market and captured some mind share, and rightfully so. And Commvault has responded effectively. It continues to be a very innovative company.”

Customers have so much value locked up in their enterprise data protection platforms, Waggoner said.

“How do we tap into the extreme value that’s potentially there?” he said. “We’ve taught for years that the challenge with AI is still how to get the data in a usable form. How do we get access to it. How do we sanitize it. How do we get consistency. How do we do all those things to make that data accessible and usable. Commvault is making a big step. Let’s face it. The central repository for most enterprises for that potential value lays is in their data protection platform. It tends to be the most consolidated location for enterprise data, and so being able to crack that open more effectively and tie it into their AI paradigms I think is super interesting.”

At next week’s Commvault Shift conference, Commvault will be making multiple product and service introductions which Brockway declined to detail.

However, he did say that Commvault will be starting to show how several of its current technologies come together.

“We will connect some of the dots,” he said. “Everybody kind of has a fundamental basis of how some of the new things come together, and we’ll show how we start to put them into a bigger, broader context.”