From AI Experiments to Enterprise Reality: The Infrastructure Overhaul

As artificial intelligence moves beyond pilots and proof-of-concept trials, enterprises face a fundamental shift in how they build and manage their IT environments. According to Nutanix leaders Tarkan Maner and Thomas Cornely, the leap from cloud-based experimentation to production-grade deployment—spanning real workloads, thousands of users, and live business processes—demands a complete rethinking of infrastructure. Below, we explore the key questions shaping this transformation.

Why is scaling AI from pilots to production so difficult?

The journey from prototype to production is fraught with hidden complexities. Running a model for a handful of users is very different from supporting 10,000 employees. Thomas Cornely, Nutanix's EVP of product management, explains that the focus has shifted from training models and simple chatbots to building autonomous agents that execute multi-step workflows. These agents place exponentially greater demands on infrastructure—real-time performance, unpredictable load spikes, and the need to coordinate access across multiple teams and data sources. Organizations must prepare not only for scale but for the reliability and governance that production environments require.

What exactly is agentic AI and why does it matter for enterprises?

Agentic AI refers to systems that can act autonomously, making decisions and executing tasks across applications and data sets without constant human oversight. This introduces a new layer of complexity: multiple agents may run simultaneously, each with unpredictable, real-time workloads. Cornely notes that tools like OpenClaw make it easy to build agents, but enterprises must ensure those agents run on-premises with proper controls to protect sensitive data. The challenge extends beyond operation to interaction—how these agents access systems, coordinate with each other, and safeguard enterprise assets becomes critical.

How does agentic AI change the relationship between humans and machines?

Tarkan Maner, Nutanix's president and chief commercial officer, emphasizes that agentic AI is not about replacing people but augmenting them. The goal is to find a harmonious balance where human judgment, AI automation, and agent-based workflows coexist. He envisions a future where love, peace, and harmony among AI tools, robotics, and human capital leads to better outcomes for businesses and public sectors. The right vendors must provide tooling and services that optimize this balance, ensuring that AI amplifies human capability rather than undermining it.

What practical steps are enterprises taking to move AI into production?

Most organizations begin with small-scale experiments, then gradually increase the number of users and complexity of tasks. Nutanix recommends building a robust infrastructure that can handle unpredictable workloads and support multiple concurrent agents. Key steps include:

• Deploying hybrid cloud or on-prem solutions for data residency
• Implementing strong governance and access controls
• Training teams to manage agent orchestration
• Using platforms that simplify agent building and deployment

Enterprises should first pilot with a limited function, measure performance, and then expand—always keeping security and compliance at the forefront.

How do data governance and security change with autonomous AI agents?

With agents acting autonomously, the attack surface expands. Agents can access multiple data sources and perform actions that might violate data policies if not properly constrained. Cornely warns that enterprises must have the right constructs—like role-based access, audit trails, and sandboxing—to prevent rogue behavior. Running agents on-premises gives businesses greater control over their data, which is especially crucial for regulated industries like banking and healthcare. Governance also means defining clear boundaries for agent decision-making and ensuring human oversight remains possible when needed.

What industries are most impacted by this shift to production AI?

According to Maner, the change is universal. Regulated industries such as banking, healthcare, government, and education face strict compliance requirements, making on-prem or hybrid infrastructure a necessity. Non-regulated sectors like manufacturing and retail also benefit from scaling AI—whether for supply chain optimization, customer service agents, or predictive maintenance. Every vertical must adapt its infrastructure to support real-time, autonomous AI without compromising security or performance. The pressure is especially high in environments where downtime is not an option.

How can enterprises prepare their infrastructure for agentic AI at scale?

Preparation involves more than adding GPUs. Nutanix suggests a platform approach that unifies compute, storage, and networking with built-in security and AI orchestration. Infrastructure must be elastic to handle unpredictable agent workloads, and it should support both cloud and on-prem deployments to meet data sovereignty needs. Companies should also invest in AI-ready operations, including monitoring tools that can track agent behavior and resource consumption. Finally, cross-team collaboration—between IT, data science, and security—is essential to ensure the infrastructure evolves alongside the AI strategy.