NetApp Charts Path for AI-Driven Storage: Guardrails Over Autonomy

The rise of AI is reshaping how enterprises think about their workloads, increasingly framing them as fundamentally data-centric challenges. The critical question facing organizations is no longer simply where their information resides, but rather who—or what—should be permitted to act upon it. Data governance has become the proving ground for whether companies can safely delegate infrastructure decisions to autonomous systems.

Enterprise information now sprawls across multiple environments: traditional on-premises data centers, public cloud platforms and neoclouds. Storage vendors are moving quickly to connect this distributed infrastructure to production AI workloads. However, fragmenting storage across multiple isolated systems is counterproductive, according to Sandeep Singh, senior vice president and general manager of enterprise storage at NetApp Inc. Singh advocates instead for a unified data foundation that maintains consistent behavior across all deployment locations.

We know that the customers are going to use a variety of locations for the right fit for them. What becomes fundamentally important for customers is to have a consistent data infrastructure strategy that underpins it all so that they can get a consistent set of data capabilities combined with a consistent operational experience across the various environments.

Sandeep Singh, NetApp

Singh and Helen Yu, founder and chief executive officer of Tigon Advisory Corp., discussed these challenges at NetApp INSIGHT, appearing on theCUBE, SiliconANGLE Media's livestreaming platform. Their conversation, moderated by theCUBE Research's Christophe Bertrand and co-host Rebecca Knight, centered on why data consistency, clear lines of accountability and policy safeguards must precede any handoff of infrastructure management to AI agents.

Data governance moves into the autonomous control plane

The challenge extends beyond technical architecture into organizational structure. Yu, who advises enterprises on AI implementation, identifies a fundamental leadership gap. When teams lack reliable access to consistent, trustworthy data through official channels, they circumvent formal systems—a pattern that spawns shadow AI and shadow IT environments.

It's not a storage problem. It's a leadership and accountability issue. When teams don't have access to consistent data, trusted data consistently, they actually stop making decisions. They start making workarounds.

Helen Yu, Tigon Advisory Corp.

NetApp's approach pairs a consistent data plane with a unified control plane that accommodates both autonomous agents and human operators. The company's hybrid cloud management tools, accessible through NetApp Console, enable customers to establish policies and operational boundaries across their entire infrastructure fleet. Agents then maintain infrastructure operations within those defined limits. In a demonstration, agents detected a performance issue during off-hours and implemented corrective measures autonomously, preventing the need to alert an engineer.

Agents can in real time implement the necessary quality of service rules to ensure there's compliance to the boundary conditions there. And that way, no human is woken up. The right outcome is achieved and there's no noisy neighbor effect. And there's an audit trail presented to the humans for going and following up.

Sandeep Singh, NetApp

Effective governance of this model hinges on accountability mechanisms as much as automation capabilities. According to Yu, simply allowing an algorithm to make decisions does not constitute a governance framework. Leaders should evaluate automation success by measuring its impact on business outcomes rather than merely counting tasks removed from human hands. Yu recommends extending traditional data governance practices to include machines by applying the RACI framework—defining who is responsible, accountable, consulted and informed.

https://www.youtube.com/embed/Q7NF25MbJ6o?start=1&feature=oembed

Now you need to incorporate the machine or agents into your RACI chart. Who owns what data? Who has access to what data? And then when you're going to make an exception, who has the right to override the exception?

Helen Yu, Tigon Advisory Corp.