Revolutionizing data protection workflows with NVIDIA AgentIQ and Cohesity’s Operational Insight Policy Agent

In today’s fast-paced, data-driven world, protecting and managing vast volumes of information has never been more critical. With enterprises grappling with an explosion of data and increasingly complex IT infrastructures, it’s clear that traditional data protection methods can no longer keep up.

At NVIDIA GTC 2025, we’re excited to introduce an innovative concept designed to revolutionize data protection: Cohesity’s Operational Insight Policy Agent, powered by NVIDIA’s cutting-edge AgentIQ toolkit, included as part of the “AI-Q’ NVIDIA Blueprint for building agentic systems. This groundbreaking technology promises to transform how organizations handle backup and recovery policies, making data protection smarter, faster, and more efficient. Say goodbye to manual intervention and hello to an automated, intelligent, and seamless approach to safeguarding your most valuable asset—your data.

Let’s examine how this powerful combination of AI-driven automation and intelligent orchestration could reshape your data management strategy.

Why we need a new kind of data protection workflow

As enterprises scale, they’re grappling with core challenges:

1. Exploding data growth: Petabytes of information is spread across on-premises, cloud, and edge environments—making it difficult to ensure comprehensive protection for all assets.
2. Complex infrastructure: Heterogeneous environments with diverse workloads require specialized backup schedules, queries, and retention policies.
3. Operational overhead: Manually managing assets, policies, and backups is tedious and prone to error, especially across thousands of systems or applications.
4. Fragmented visibility: Existing policy management tools often fail to present a single pane of glass, making it challenging to quickly identify unprotected assets or evaluate whether existing policies align with new business requirements.

Cohesity’s Operational Insight Agent, possibly built using the NVIDIA AgentIQ toolkit, would aim to directly address these challenges. By offering an adaptable, intelligent, and workflow-driven system, we would automatically recommend, create, and activate backup policies, drastically simplifying data protection for modern IT operations.

What is NVIDIA AgentIQ?

NVIDIA AgentIQ is an open-source toolkit designed to connect, accelerate, and optimize agentic systems in modular, scalable environments. It allows for building “agents,” independent modules or microservices, that communicate through well-defined interfaces. Each agent performs a specialized role within a broader workflow, making it easy to update, swap out, or add new functionality without overhauling the entire system.

For Cohesity, NVIDIA AgentIQ offers the following benefits:

• Modular design: Agents can be developed, tested, and deployed independently, reducing development complexity.
• Intelligent orchestration: Complex decision-making can be distributed across agents or centralized in a “supervisor,” depending on the architecture, allowing for flexible design patterns.
• Rapid iteration: NVIDIA AgentIQ agent-based approach allows new features or improvements to seamlessly integrate without disrupting existing workflows.

Cohesity Operational Insight Policy Agent

Our Operational Insight Agent would be the brain behind Cohesity’s new automated policy workflow. It would use NVIDIA AgentIQ to connect multiple specialized sub-agents into a cohesive pipeline. Each sub-agent would be responsible for a key function in the policy management lifecycle, from discovering unprotected assets to recommending, creating, and activating backup policies.

Key pain points addressed:

• Automated discovery: Use AI-driven queries to find unprotected or recently discovered assets across your environment.
• Policy recommendations: Dynamically generate recommended backup policies based on your asset types, risk profiles, and compliance requirements.
• Simplified policy creation: Automatically create or modify backup policies with minimal manual intervention.
• Activation and validation: Ensure newly generated policies are correctly activated, validated, and aligned with user intent and compliance guidelines.

Two architectural approaches: Supervisor vs. Decentralized

When investigating the integration of these agents into an AgentIQ-based ecosystem, Cohesity explored two main approaches to handle workflow routing and supervision.

Approach #1: Hierarchical agents under a Supervisor

1. Overview:

• Retains an overarching “Supervisor” agent that delegates tasks to specialized agents.
• Allows other agents to call subordinate agents as needed without major structural changes.

2. Pros:

• Minimal disruption to existing policy workflows.
• Reduced complexity for the Supervisor agent by delegating certain tasks to local sub-agents.
• Simplified debugging: Clear lines of communication make it easy to diagnose issues.

3. Cons:

• In certain flows (e.g., existing policy recommendations), some agents (like Creator or Activator) are never used, adding overhead.
• The Supervisor’s prompt can still grow in complexity since it must handle numerous scenarios.
• Over time, the hierarchical approach can be cumbersome if more specialized agents are added.

Approach #2: Decentralized workflow without a Supervisor
1. Overview:

• Eliminates the central Supervisor agent.
• Each agent becomes “smarter,” independently determining the next workflow step.

2. Pros:

• Simplified routing: There’s no need for a single decision-making point, which can reduce latency.
• Modular: Each agent can operate independently, making the system easier to maintain and update.
• Scalability: More agents can be added without overcomplicating a central routing mechanism.

3. Cons:

• Larger refactoring effort: Transitioning from a supervisor-based design is non-trivial.
• Increased individual agent complexity: Each agent must be more aware of the overall flow to make correct decisions.

Choosing the right approach

For organizations already running a Supervisor-based architecture, Approach #1 would offer a more seamless transition without overhauling existing workflows. It could quickly integrate any potential new specialized agents, like the Operational Insight Policy Agent, under a familiar structure.

However, for greenfield projects or those looking for maximum agility and minimal routing overhead, Approach #2 could unlock greater efficiency. By distributing workflow decisions across specialized agents, businesses reduce the risk of a single bottleneck while gaining a system that scales elegantly with additional use cases.

How NVIDIA AgentIQ makes it all easy.

The NVIDIA AgentIQ toolkit is a natural fit for these approaches because:

• Agent reusability: Modules (agents) can be reused across multiple workflows or expanded for new functionalities without rewriting everything.
• Inter-agent communication: NVIDIA AgentIQ provides robust connectors for passing data among agents—eliminating friction when evolving from Approach #1 to Approach #2 or vice versa.
• Adaptive decision-making: Built-in intelligence layers allow each agent to make local decisions or escalate to a higher-level orchestrator, depending on your chosen design.

Real-world use case: Automated backup policy creation

Imagine a large enterprise with thousands of VMs, databases, and containers. The automated backup policy creation process could be:

1. The Operational Insight Policy agent uses the Workload agent to detect recently added assets that have no backup coverage.
2. The Collector agent confirms if those assets require coverage
3. The Recommender agent crafts a brand-new policy or identifies an existing policy match.
4. Either the Creator or Activator agent seamlessly implements and activates these policies, all without requiring constant manual oversight.

The result? Fewer lost assets, better compliance, and more efficient use of IT resources.

The road ahead

As data volumes explode and organizations demand tighter SLAs, NVIDIA AgentIQ paired with solutions like the Cohesity Operational Insight Policy Agent, are set to play a major role in the future of data protection. Our two architectural approaches, Supervisor-based or Decentralized, could offer flexible paths to integrate next-gen policy workflows without sacrificing performance or ease of management.

We’re excited to see how customers and partners will innovate on top of Cohesity’s platform with NVIDIA AgentIQ. Whether you choose a gradual evolution from a Supervisor agent or a bold move toward a decentralized model, the Operational Insight Agent will stand ready to simplify policy management, boost resiliency, and help you sleep a little easier at night.

Join us at GTC 2025

To learn more about this integration and see a live demo of our Operational Insight Agent in action, join us at our NVIDIA GTC 2025 session or drop by our booth #2012. We’ll showcase how NVIDIA AgentIQ agentic power converges with Cohesity’s data management platform to drive truly intelligent, automated policy workflows—no matter how large or complex your environment becomes.

Stay tuned for future updates, and visit Cohesity’s booth #2012 at GTC 2025 to experience the next wave in intelligent data protection.