What to watch for as enterprises move from AI experimentation to autonomous operations

In the past three years, discussions around enterprise artificial intelligence (AI) have centered on the capabilities of AI models to generate insights. However, the next three years will focus on what autonomous AI systems are permitted to do. During the Bloomberg Tech Event: Building the Agentic Enterprise in Boston on October 28, industry leaders from various sectors will discuss the challenges and opportunities of deploying agentic AI, including chief information officers, chief security officers, and heads of data, analytics, and AI.

The transition from experimental AI usage to operational autonomous AI systems is a complex one. While vendors may tout the idea of autonomous operations, ultimately it is the operators who determine when and how to implement these technologies. The shift from viewing infrastructure as merely plumbing to a strategic component of AI has been highlighted by HPE's Networking Investor Day.

This shift reflects the growing expectation for AI systems to operate autonomously, requiring infrastructure capable of running AI systems independently.

CIOs and CTOs should pay attention to several key developments in this evolving AI landscape. One critical area is determining the boundaries between fully automated actions, actions requiring human approval, and actions that remain manual. Advanced companies do not adopt autonomous operations as a binary switch, but rather evolve gradually, introducing autonomous actions one category at a time.

Another important aspect of autonomous AI systems is the shift in measuring infrastructure health. Instead of solely focusing on infrastructure health, organizations must begin measuring user and business outcomes. For instance, a network may be fully available, but if it fails to deliver satisfactory user experiences, such as poor video calls or stalled AI workloads, it has not fulfilled its purpose. This change in perspective is crucial for the successful implementation of autonomous operations.

At the event, speakers will discuss their perspectives on measuring success in the agentic era. Key metrics may include hours saved by engineers, incidents avoided from reaching a ticket queue, and faster time to value for new services, rather than merely reducing headcount. The networks must be optimized to handle the unique demands of AI traffic, which requires continuous connectivity and balanced upstream and downstream traffic.

As AI agents become a new class of users on the network, the network must play an identity and policy layer role, not just a transport layer.

Security concerns are also paramount as AI agents require access to various systems, data, and each other, creating a growing population of nonhuman identities. Traditional zero-trust models are inadequate in this context, and the emergence of AI-powered vulnerability scanners, like Anthropic's Claude Mythos, has heightened the need for modernizing network infrastructure to protect against AI-generated threats.

As the network becomes a sensor and a defense mechanism against AI threats, CISOs must navigate the challenges of managing the transition to autonomous operations while ensuring security and risk management.