Data protection provider Commvault Systems Inc. announced the launch of its Recovery Studio during the SHIFT 2026 customer event, a new service designed to automate the restoration of entire business applications following an outage or cyberattack.
In addition to Recovery Studio, Commvault made four other significant announcements regarding recovery solutions tailored for artificial intelligence agents, as well as collaborations with Databricks Inc., Nvidia Corp., and Zscaler Inc.
The company is expanding its AgentRecover tool, which is used to mitigate damage caused by AI agents, to support agents built on Microsoft Corp.’s Foundry and Amazon Bedrock AgentCore from Amazon Web Services Inc. The integration of Databricks and Nvidia technologies will leverage governed enterprise data from Commvault’s Data Activate service. Furthermore, a strategic partnership with Zscaler focuses on enhancing AI security.
Recovery Studio aims to streamline the recovery process, eliminating the traditional manual approach that restores applications piece by piece. It recognizes that simply having the data is insufficient for reactivating an application; without its identities, configurations, and underlying infrastructure, an application may remain offline well beyond the point when its files are restored.
For large enterprises, the recovery challenge can be daunting, as they may need to sequentially restore hundreds of applications and thousands of identities. Commvault’s research indicates that average recovery time using conventional methods is approximately 28 days, as reported in their research from the previous year.
To overcome this complexity, Recovery Studio employs automated runbooks that systematically rebuild and validate each application component. According to Commvault, these runbooks also restore identity systems and cloud configurations, including firewall rules. Threat scans can be integrated within a runbook for supported resources, guiding decisions about whether a service is ready to resume production. All recovery operations—be it operational, disaster, or cyber recovery—will be managed from a unified platform.
Pranay Ahlawat, Chief Technology and AI Officer at Commvault, described the launch as “the DevOps moment for recovery,” noting that a single orchestrated flow can replace the reliance on spreadsheets and war rooms commonly used by recovery teams today.
The enhancements to AgentRecover extend the company’s protection capabilities to the environments where businesses develop their production agents. In September, Commvault introduced options to safeguard agent configurations, such as those used by Anthropic PBC’s Claude Code on employee laptops.
Phil Goodwin, Research Vice President at International Data Corp., noted that traditional recovery methodologies assume a static dataset at a specific timestamp. AI agents disrupt this premise, as the agent, its data, and associated resources need to be reinstated “at machine speed,” he explained.
On the Bedrock AgentCore and Foundry platforms, the software automatically discovers each agent and maps the data, applications, and infrastructure involved. In the event an agent misbehaves, its activity log can reveal all systems it has potentially impacted across various platforms.
The restoration process targets the specific agent and any alterations it has made. Additionally, customers can utilize AgentRecover alongside Commvault’s identity recovery and change auditing tools to trace transactions associated with agent activities.
Both the Databricks and Nvidia collaborations build upon Data Activate, which transforms backup copies managed by Commvault into datasets suitable for analytics and AI. A significant portion of corporate data remains unstructured and dispersed across various systems, with research from BARC, published by DataHub in June, indicating that 70% of data and AI leaders reported less than half of their unstructured data was accessible or suitable for AI applications.
Commvault Data Activate allows data to be shared with Databricks under Commvault’s governance, delivering it as governed tables within Unity Catalog, maintaining ownership, access permissions, and lineage information.
Ongoing updates ensure that as source content evolves or is deleted, customers are not required to rebuild connections to all original sources. Alan Atkinson, Chief Business Development Officer at Commvault, stated that the goal is to “transform protected data into an active business asset.”
Nvidia’s contribution, delivered as a reference design for software-as-a-service and software deployments, focuses on data retrieval. The combination of Nvidia NIM microservices and Nemotron Retriever, together with the Nvidia cuVS library utilized with Apache Solr, enables easier extraction and organization of Commvault-protected content for applications and agents. Data Activate continuously updates this content as source files change, retention periods expire, or permissions are modified.
In terms of security, Commvault and Zscaler have identified AI discovery, runtime governance, and agent detection and recovery as initial collaborative areas. However, they did not announce a joint product or timeline. Zscaler’s Security for AI line emphasizes monitoring how employees and agents utilize AI tools. Dhawal Sharma, Zscaler’s Executive Vice President of AI Security and Strategic Initiatives, emphasized the necessity for organizations to be “prepared to recover quickly if something goes wrong.”
The features of Recovery Studio, along with AgentRecover’s support for Amazon Bedrock AgentCore and Databricks integration, are expected to enter early access in December, with the last of these offerings available in select Microsoft Azure regions. Meanwhile, support for Foundry is anticipated to be available sooner, with early access scheduled for this month.
The Nvidia reference design is projected for general availability in 2027, following an early release in December. Pricing for Recovery Studio will be based on the amount of data configured for orchestrated recovery, with general availability expected in the coming months.
A virtual edition of SHIFT is scheduled for November 18.