Cohesity has introduced Agent Resilience, a new Cohesity Data Cloud capability designed to discover, protect and recover the infrastructure behind enterprise AI agents. Initially supporting Amazon Bedrock, the technology extends conventional cyber resilience to agent memory, configuration and connected data resources as enterprises deploy AI systems capable of taking actions without continuous human review.
Cohesity is extending its cyber resilience platform into the emerging market for AI agent protection with Cohesity Agent Resilience, a capability designed to recover both the state of an AI agent and the enterprise resources it affects.
Announced at Cohesity Catalyst 2026, the technology addresses a problem that becomes more consequential as AI systems move from answering questions to executing multi-step workflows. Agents can query databases, modify records, trigger applications and use privileged credentials. If an agent’s memory, configuration or operating context becomes corrupted, simply restoring the underlying database or application may not restore the agent itself to a trusted state.
Cohesity Agent Resilience is designed to provide that additional recovery layer. The company says the capability can discover agent dependencies, protect agent state and restore supported agents to a known-good point in time. It is currently available to select customers, with general availability targeted for the end of 2026.
The initial release supports Amazon Bedrock AgentCore and Amazon Bedrock Agents. Cohesity says Microsoft and Google agent platforms are planned for future support.
The technology is built around what Cohesity calls Agent Topology, which maps relationships between an agent, its memory stores, configuration and connected data resources. That provides IT and security teams with a view of the dependencies that may need to be recovered when an agent-driven incident occurs.
The protection model has two sides. First, Agent Resilience protects the agent stack itself, including memory and configuration, using point-in-time recovery, immutable backups and Cohesity’s existing recovery architecture. Second, it protects the databases, file systems and other services that agents interact with.
That distinction is becoming important as enterprises adopt agentic AI. An AI assistant generally recommends an action for a person to approve. An autonomous agent can execute the action itself, creating a larger operational blast radius when something goes wrong.
Gartner projects that up to 40% of enterprise applications will include integrated task-specific AI agents by the end of 2026, compared with less than 5% in 2025. Gartner has also warned that by 2027, 40% of enterprises could demote or decommission autonomous AI agents because of governance failures identified after production incidents.
Those projections help explain the emerging focus on resilience alongside agent governance and observability. Monitoring can identify unusual behavior, while access controls can restrict what an agent is allowed to do. Recovery provides a different capability: returning the affected system or agent to a trusted state after an unwanted action has already occurred.
Cohesity is using this distinction as the foundation for its broader Autonomous Cyber Resilience strategy. The company describes the initiative as a future model for automating its five-step cyber resilience framework through agentic workflows: protecting data, identity, applications and agents; maintaining recoverability; remediating cyber and AI threats; practicing application recovery; and optimizing data and AI risk posture.
The company says administrators would eventually be able to define resilience objectives through Cohesity Copilot, including recovery time objectives and recovery point objectives, threat-scanning requirements and recovery-testing requirements. The Data Cloud would then assess workloads, recommend protection and recovery strategies, and present those recommendations for approval before execution.
During an incident, the proposed architecture would automate portions of impact assessment, threat identification and isolated recovery preparation.
Cohesity has already introduced one component of that strategy. Customers using Cohesity Data Cloud Enterprise Edition and Cohesity Data Security Posture Management can automate protection for newly discovered sensitive data that is not already covered by protection policies. The workflow continuously discovers and classifies sensitive information, checks its protection status and can trigger policy-based protection actions.
The company’s approach builds on Cohesity RecoveryAgent, which orchestrates elements of incident response and recovery. Cohesity also plans to expand Maestro later in 2026 to connect protection, response and recovery capabilities with external AI tools including Claude, ChatGPT and Gemini, alongside Cohesity Helios.
The company has been building toward this positioning throughout 2026. In March, Cohesity and ServiceNow announced a partnership focused on resilience for autonomous AI agents, while Cohesity separately announced integrations with Datadog combining observability with automated data recovery.
That ecosystem approach places Cohesity alongside a growing category of security, observability, governance and data-protection vendors trying to address the operational risks created by agentic AI. The market is shifting from securing AI models and applications alone toward protecting the complete environment in which agents operate.
Cohesity is also attempting to build an education layer around the issue. Its new AI Resilience Academy begins with a free, self-paced Foundations of AI Resilience with Cohesity course. The 25-to-30-minute course focuses on the difference between assistants and agents, the additional risks created when AI can take actions, and the roles of Cohesity’s AI resilience technologies.
The larger technology trend is clear: as AI agents gain access to business systems, resilience increasingly needs to cover not just the data those systems contain but the state and context that determine how autonomous software behaves.
Cohesity’s Agent Resilience is an early attempt to formalize that requirement as a dedicated recovery capability. Its success will depend on how broadly agent platforms are supported and how effectively organizations integrate recovery with existing AI governance, identity, security and observability systems.
For enterprises moving agentic AI into production, however, the underlying question is becoming harder to avoid: if an autonomous system can change the business environment, the organization also needs a reliable way to undo those changes.
Market Landscape
Enterprise AI is moving toward systems that can execute tasks across multiple applications rather than simply assist individual users. Gartner estimates that up to 40% of enterprise applications will include task-specific agents by the end of 2026, while separately predicting that 40% of enterprises could demote or decommission autonomous agents by 2027 because of governance gaps.
This creates a new layer of enterprise AI infrastructure around agent governance, identity, observability, data protection and recovery. Traditional backup protects data and applications, but agentic systems introduce additional state in the form of memory, prompts, configurations, permissions, guardrails and workflow context.
Cohesity is competing within this broader AI security and resilience landscape alongside observability, cybersecurity, cloud infrastructure and AI platform vendors. Its differentiation is the emphasis on restoring an agent and its dependencies to a trusted state after an incident rather than focusing exclusively on prevention or monitoring.
Top Insights
- Cohesity Agent Resilience protects AI agent state and connected enterprise resources, adding recovery capabilities to existing agent governance and observability layers.
- The initial release supports Amazon Bedrock AgentCore and Amazon Bedrock Agents, with Microsoft and Google platforms planned.
- Agent Topology maps agents, memory, configurations and connected resources to help identify dependencies and protection gaps.
- Cohesity is developing Autonomous Cyber Resilience around agentic workflows for protection, threat response, recovery testing and risk optimization.
- Gartner expects task-specific agents to become part of up to 40% of enterprise applications by the end of 2026.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
