A Cisco executive told a U.S. Senate subcommittee that artificial intelligence is fundamentally changing how enterprise and service provider networks are built, operated, and secured. Bob Everson, chief architect of provider mobility at Cisco, testified before the Senate Subcommittee on Telecommunications and Media on July 30 at a hearing titled “Intelligent Networks: Powering Artificial Intelligence and Transforming Communications.” The hearing was designed to examine how widespread AI adoption has forced networks to evolve, requiring more capacity and complex designs so that AI workloads can run efficiently.

Senator Deb Fischer, a Nebraska Republican who chairs the subcommittee, said in her opening statement that the panel would consider how government, providers, and other industries are responding to AI-driven demand. She noted that private companies have invested hundreds of billions of dollars in network deployment in recent years, and that federal broadband programs have contributed billions more for targeted deployment and maintenance. Other witnesses included representatives from U.S. Telecom, Vanderbilt University, and the Nebraska Public Service Commission.

AI is changing network traffic behavior

Everson said that AI is not simply increasing the volume of network traffic; it is also changing the behavior of that traffic. Cisco measured a fourfold increase in AI inference traffic over an eight-month period, he said. Traditional networks have been optimized for downstream content delivery, but AI workloads are far more two-way and uplink-intensive. Prompts, context, sensor data, and agent activity travel back toward AI models, and the resulting connections remain active longer than conventional web transactions.

AI agents amplify those effects by operating at software speed. In Cisco testing, an agent generated 450 percent more traffic than a person performing the same task, and roughly 70 percent of that additional traffic was inference. This shift has direct consequences for campus, branch, and service provider networks as they try to keep up with new usage patterns.

Everson cited customer reports showing a 34 percent increase in campus and branch traffic tied to AI workloads over the past 12 months, with expectations of a 96 percent increase in the coming year. Half of enterprise customers report that AI demand is concentrated on their Wi-Fi networks, and 73 percent of organizations already face or expect to face campus and branch capacity limitations within the next 24 months. These pressures stem from increases in east-west traffic, latency-sensitive traffic, and continuous automated AI traffic.

He also noted that while most AI workloads to date have relied on foundation models running in central infrastructure, enterprises are increasingly deploying small language models, open-source models, and specialized vision and voice models that can be distributed throughout the network. That trend reinforces the value of the Federal Communications Commission’s 2020 decision to authorize the full 6 GHz band for unlicensed Wi-Fi use, he said.

Infrastructure and edge computing pressures

In his prepared remarks, Everson identified several areas affected by AI. The first is infrastructure. AI is driving a shift toward edge computing, and service providers must account for “AI-native” traffic profiles because of technical considerations, cost, and concerns about data sovereignty and security.

The technical case for edge computing is especially clear in physical AI applications. Robotics, autonomous vehicles, and industrial automation may require sub-millisecond decision-making. If an autonomous robot sends data to a central cloud and waits for a response, the round-trip latency could be too high for safe real-time operation, Everson said. Placing compute capacity closer to the network edge helps reduce that delay.

Cost is another driver. AI operations generate massive amounts of data. High-definition video analytics for public safety, for example, can produce terabytes of data every day. Backhauling that data to a central cloud is prohibitively expensive and can create severe network congestion. Distributing compute and inference capabilities across the network can reduce the amount of data that must traverse long-distance links.

Data sovereignty and security also matter, Everson said. Enterprises and government agencies are increasingly concerned about moving sensitive information across the public internet to a third-party cloud provider. Regulatory and security requirements often make it difficult to send certain data outside a particular jurisdiction or network boundary. AI-native architectures that support local processing can help address those concerns.

AI-driven network management and self-healing

While AI workloads create challenges, Everson argued that networks can also leverage AI to improve performance, reliability, and security. Agentic AI, he said, will change the nature of network traffic but also give operators new tools to work at machine speed and deliver better performance and efficiency.

Cisco’s AI-native tools, he said, enable the network to reroute traffic, adjust capacity, or reconfigure nodes when the system detects performance degradation or an impending hardware failure. This self-healing capability increases uptime and reliability for mission-critical services. Everson said this approach is distinct from traditional network management, which often relies on manual intervention after a problem occurs.

The talent gap in networking is another major challenge. Managing increasingly complex, software-defined networks requires specialized skills. Everson said that AgenticOps can automate repetitive, low-value tasks such as ticket resolution, configuration updates, and routine maintenance. These tools lower the barrier to entry and allow junior analysts to ramp up quickly. By automating those tasks, AI-enabled platforms free network engineers to focus on higher-level architectural strategy and innovation. Cybersecurity analysts can also spend more time on strategic threat hunting and detection engineering.

Moving toward AI-native platforms

Everson said networks are moving from being simple pipes to becoming the fabric of intelligent connectivity. As network operators move compute toward the edge, such as at cell sites, they will be able to run applications directly from the network. This evolution sets the stage for new capabilities like Integrated Sensing and Communication, or ISAC, which combines wireless communications and radio-frequency sensing to detect objects’ position and path using radio waves that reflect off them.

ISAC can detect intrusion even in low-light conditions, through smoke, or around obstructions where traditional video analytics might fail. Everson noted that the technology has already been prototyped and demonstrated, and it holds promise for autonomous systems, robotics, AI-driven smart facilities, and public safety.

Policy recommendations for U.S. leadership

Everson closed with three suggestions for the committee. First, he urged Congress to accelerate the U.S. AI-native stack. Cisco is investing in AI-native networking, bringing new capabilities to 5G-Advanced and laying the groundwork for 6G. He highlighted AI-WIN, a collaboration among Cisco, NVIDIA, MITRE, Orion Development Company, Booz Allen, and T-Mobile, which aims to bring AI, compute, and wireless together to create a secure, American-led path from 5G-Advanced to AI-native 6G. He encouraged Congress to focus on areas where the United States has strategic leadership, including compute, core networking, and applications.

Second, he called for modernizing permitting and infrastructure processes. As computing becomes more distributed, permitting rules must enable rapid and responsible deployment. In particular, he said the committee should account for the evolving costs of AI-ready networks when considering the future of the Universal Service Fund, so that rural and urban communities can share in the benefits.

Third, he emphasized the need for a balanced spectrum policy. The 800 megahertz of licensed spectrum recently made available by Congress is essential for high-capacity, high-uplink connectivity, he said. The FCC’s authorization of the 6 GHz band for unlicensed use is equally important for meeting enterprise demand. A dependable pipeline of both licensed and unlicensed spectrum is foundational to American leadership, he said.

Source: Network World News