The Machine Speed Imperative for Enterprises

Enterprise technology has traditionally operated at human speed. While many repetitive tasks have long been handled by automation, the speed of critical decisions, workflows and exception handling have remained constrained by human capacity, as people still monitor systems, interpret information, make judgments and direct machines.

The introduction of generative AI accelerated the first part of that loop, helping people analyze information faster and produce outputs more efficiently.

The new era of agentic AI completely changes the game. AI agents can now coordinate workflows, make decisions and execute tasks independently and autonomously.

Researchers from Carnegie Mellon and Stanford conducted a 2025 study comparing human workers and AI agents across data analysis, engineering, computation, writing and design. The study found that agents completed tasks at least 88.3% faster, with the caveat that the agents’ work was of lower quality. Yet, since that study was done, AI capabilities and output quality have only continued to progress.

This is ushering in a new era, where enterprises can now operate at machine speed. 

The capabilities of agentic AI are putting CIOs under increasing pressure to harness automation for greater speed, efficiency, competitiveness and resilience.

James Robertson says that CIOs must take this moment seriously. A technology executive and strategist in the office of the CTO at HPE, Robertson has spent more than 30 years leading large-scale infrastructure, networking and technology transformations.

“The biggest risk for the CIO is to stay the course and operate at human speed while rivals move at machine speed,” says Robertson. “That’s a recipe for disaster.”

CIOs are already shifting budgets to meet AI demands. From January 2025 to January 2026, Bloomberg Intelligence found that spending momentum moved beyond AI models and into the infrastructure needed to run them.

But simply adding AI capacity isn’t enough. Companies also need to modernize their underlying compute, storage and networking infrastructure so it is fast enough to keep pace with machine-speed systems.

CIOs may be tempted to treat agentic AI as just another software project, says Robertson: Build a model, connect it to enterprise data and let it go to work. But often, autonomous systems can accelerate the decision-making layer faster than the underlying enterprise infrastructure can keep up.

An AI agent, for instance, may approve an order in seconds. But if the inventory database updates only once an hour, or the API connecting it to the ERP system can’t handle the increased load, the company hasn’t eliminated its bottlenecks — it has simply automated them.

Robertson and his team at HPE have developed a framework to help customers build a truly agentic enterprise. The framework rests on three ideas: Build, Govern and Operate. Enterprises need to build a secure AI-native foundation; govern how autonomous systems access data and take action; and operate the entire environment as one intelligent system.

“I wouldn’t build the roof of a house before I built the foundation and the walls,” says Robertson. “I wouldn’t take AI and deploy it without understanding the performance, governance and security requirements that I need.”

Building the foundation for agentic AI doesn’t mean ripping and replacing everything that already works. It means modernizing systems for a world where AI increasingly acts autonomously and at machine speed. 

Data is the foundation of AI, and the network moves data across the enterprise and connects the applications, users, agents, policies and infrastructure that enable autonomous action. 

But a machine-speed enterprise cannot rely on a network operated entirely at human speed. For AI agents to continuously interpret information, coordinate workflows and take immediate action, the network connecting them must also be intelligent and responsive. Otherwise, the network risks becoming the next human-controlled bottleneck in an increasingly autonomous operating model.

AI is not simply creating more demand for network upgrades; it is also changing the role of the network itself. Networks are evolving from primarily transporting data to understanding user experiences, identifying emerging issues, applying security policies and initiating actions for ongoing optimization.

Robertson says that this expanded role is the idea behind the HPE Self-Driving Network — HPE’s network architecture that applies AI, automation and extensive insights to user and application experiences, and shifts IT from reactive troubleshooting to proactive and increasingly autonomous operations.

It can identify and diagnose issues, recommend or take corrective actions and optimize experiences with little or no manual intervention.

HPE’s Marvis AI Assistant, the networking industry’s first conversational AI assistant that provides proactive and predictive actions, illustrates how this transition is already taking shape.

The HPE Self-Driving Network is therefore a firm foundation of the operating model for an agentic enterprise, says Robertson: It helps organizations build the connectivity AI requires and govern how autonomous systems securely interact with users and resources, and it supports complex environments with the speed, visibility and resilience machine-speed business demands.

As an enterprise allows AI more autonomy, its network requires greater intelligence. The long-term ambition is an environment in which both AI agents and the network supporting them can operate within established policies, sense changing conditions and respond at the speed and scale the enterprise requires.

Real-world deployments show how AI-native networking is increasingly moving from concept to operational reality. With HPE, ServiceNow has reduced user-opened network support tickets by over 90%, and Uber’s AI-native network, which deploys the HPE Networking AI-Native SD-WAN solution,has cut infrastructure costs in half and reduced network latency by more than 10x, from 100ms to as little as 8ms.

The more decisions agents make, the more important it becomes to define what they’re allowed to do. An agent should have a clearly defined role, and the infrastructure should enforce those boundaries.

“Having autonomous agents doesn’t mean that they’re not controlled,” Robertson says. “It means they are continuously governed by an overarching framework that’s helping to make sure they don’t step out of bounds.”

That makes zero-trust principles increasingly relevant to autonomous software. Strong identity protocols, scoped authorization, least privilege rules, continuous verification, visibility and auditability provide guardrails that let agents act within clearly defined boundaries.

These guardrails must accompany agents wherever they interact with applications, data, users and infrastructure. Because the network hosts these points of connection, it plays an important role in translating governance intent into operational policy, visibility and enforcement across the enterprise. The self-driving network is not autonomous without boundaries; it is designed to become more autonomous within them.

The objective isn’t to eliminate human review and oversight, but to reduce the need for humans to intervene in routine decisions. People should continue to establish policies, handle exceptions and remain accountable for consequential decisions.

As AI systems move from assisting employees to acting autonomously, governance can’t be an after-the-fact compliance exercise. HPE, for example, subjects AI projects to risk assessments against principles covering privacy and security, human review, bias, accountability and robustness, with safeguards and escalation built into the process.

To operate effectively as an agentic enterprise, organizations need to define how intent, decisions, exceptions and accountability flow through an intelligent system designed with security and trust in mind.

The payoff is an enterprise that can quickly sense changes, make decisions and act without routing every step through a human bottleneck. Decisions move faster, routine problems can be addressed earlier and IT teams can spend less time monitoring systems and more time improving the enterprise.

But operating at machine speed is not about autonomy for its own sake. A self-driving network environment must operate within the intent, policies and accountability established by people. The advantage comes from pairing machine-speed action with systems designed for trust, resilience and human oversight.