Artificial intelligence is fundamentally reshaping the way enterprise and service provider networks are designed, operated, and secured, according to a Cisco executive who testified before a U.S. Senate subcommittee last week. Bob Everson, chief architect of provider mobility at Cisco, appeared before the Subcommittee on Telecommunications and Media, which is part of the Senate Committee on Commerce, Science, and Transportation. His testimony centered on two core questions: how AI is reshaping networks, and how networks can harness the power of AI to become more resilient and intelligent.

The July 30 hearing, titled “Intelligent Networks: Powering Artificial Intelligence and Transforming Communications,” was convened to explore how the rapid adoption of AI has impacted network infrastructure. In her opening statement, U.S. Senator Deb Fischer (R-Neb.), Chairman of the subcommittee, noted that widespread AI use has forced networks to evolve, requiring more capacity and more complex designs so that AI can run efficiently. She also highlighted the significant investments already made by private companies and federal broadband programs to support network deployment and maintenance across the country.

Everson was joined by witnesses from U.S. Telecom, Vanderbilt University, and the Nebraska Public Service Commission, providing a broad perspective on how AI is affecting both commercial and public networks. His testimony drew on Cisco’s extensive telemetry and customer data, offering concrete examples of how AI traffic patterns differ from traditional internet traffic.

AI is changing network traffic behavior

“AI is changing not only the volume of network traffic, but the behavior,” Everson told the panel. He cited Cisco measurements showing a fourfold increase in AI inference traffic over just eight months. Traditional networks have been optimized for content flowing downstream, such as video streams and web pages. AI, however, is far more two-way and uplink-intensive. Prompts, context, sensor data, and agent activity all travel back toward AI models, and the resulting connections remain active much longer than conventional web transactions.

This shift is amplified by AI agents, which operate at software speed. According to Cisco’s 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 means that enterprises planning for AI adoption must rethink their network architectures to accommodate not just higher volumes, but different traffic patterns characterized by sustained bidirectional flows and increased latency sensitivity.

Capacity strain on campus and branch networks

Everson provided specific data on how AI workloads are straining campus and branch networks, including the one that powers the Senate office building where the hearing was held. “We have already seen customers report a 34% increase in traffic tied to AI workloads over the last 12 months, and they expect to see a 96% increase this coming year,” he said.

Furthermore, half of enterprise customers report that AI demand is concentrated on their Wi-Fi networks, and 73% of organizations already face or expect to face campus and branch capacity limitations within the next 24 months. This is largely because organizations are reporting significant increases in east-west traffic, latency-sensitive traffic, and continuous automated AI traffic.

“While the large majority of AI to date has come from foundation models running on central infrastructure, we are seeing enterprises deploy more small language models, open-source models, and specialized models—such as vision and voice models—which can be distributed throughout the network,” Everson noted. He underscored the value of the Federal Communications Commission’s 2020 decision to authorize the full 6 GHz band for unlicensed Wi-Fi use, calling it a forward-thinking move to support this growing demand.

Key impacts of AI on network infrastructure

In his prepared remarks, Everson cited several areas that are directly impacted by AI. These include infrastructure, technical requirements, cost, and data sovereignty and security.

Infrastructure: AI is driving a shift toward edge computing. Service providers must consider “AI-native” traffic profiles for a variety of reasons, including technical considerations, cost, and data sovereignty and security issues. Rather than centralizing all computation in a few large data centers, networks will increasingly distribute AI processing closer to the data source.

Technical: Physical AI use cases such as robotics, autonomous vehicles, and industrial automation could require sub-millisecond decision-making. Everson explained that if an autonomous robot sends data to a central cloud and has to wait for a response, the round-trip latency could be too high for safe, real-time operation. This demands a network fabric that can support ultra-low-latency communication with localized compute resources.

Cost: AI operations generate massive amounts of data. For example, high-definition video analytics for public safety can generate terabytes of data daily. Backhauling that data to a central cloud is prohibitively expensive and creates massive network congestion. Processing data at the edge can dramatically reduce backhaul costs and improve response times.

Data sovereignty and security: Enterprises and governments are increasingly concerned about data sovereignty and security. Many customers have security or regulatory concerns about moving sensitive information across the public internet to a third-party cloud provider. By keeping data processing and storage closer to its origin, organizations can better comply with regulations and maintain control over their data.

Potential benefits of AI-driven networks

While AI workloads present challenges for network operators, Everson emphasized that there is also tremendous opportunity. “Networks will leverage AI for increased performance and resiliency,” he said. “Agentic AI will change the nature of traffic on the network, but it will also provide network operators new tools to operate at machine speed and deliver greater performance, efficiency, and security.”

One of the key concepts Everson introduced is AgenticOps, which allows the network to act as a self-healing system. Cisco’s AI-native tools enable the network to reroute traffic, adjust capacity, or reconfigure network nodes when the system detects performance degradation or an impending hardware failure. This dramatically increases uptime and reliability for mission-critical services, a crucial requirement for industries such as healthcare, finance, and public safety.

Addressing the network talent gap

Everson also spoke about one of the most significant challenges for network operators: the talent gap in managing increasingly complex, software-defined networks. AgenticOps allows operators to automate repetitive, low-value tasks such as ticket resolution, configuration updates, and routine maintenance. These tools help close the workforce talent gap by lowering the barrier to entry and allowing more junior analysts to ramp up quickly.

“By automating these tasks, Cisco’s AI-enabled platforms can free network engineers to focus on higher-level architectural strategy and innovation, and free cybersecurity analysts to dedicate more time to strategic threat hunting and detection engineering,” Everson stated.

The path toward AI-native platforms

In addition to changes in traffic patterns, Everson described how networks are moving toward AI-native platforms that will become the fabric of intelligent connectivity rather than a simple pipe. As network operators move compute toward the network edge, such as at a cell site where a tower sits, they will be able to run applications directly from the network.

One promising application is Integrated Sensing and Communication (ISAC), which combines wireless communications and radio-frequency sensing to “see” objects’ position and path using radio waves that reflect off them. Unlike optical sensors, ISAC can detect intrusion even in low-light conditions, through smoke, or around obstructions where traditional video analytics might fail. Everson noted that this technology has been prototyped and demonstrated already, and it holds great promise for autonomous systems and robotics, AI-driven smart facilities, and public safety.

Policy recommendations for the committee

In closing, Everson offered three specific suggestions for the committee to consider in future actions. First, he urged Congress to accelerate the U.S. AI-native stack. Cisco is investing across multiple dimensions of AI native networking, bringing new capabilities to 5G-Advanced today while building the foundation for 6G. One example is AI-WIN, a collaboration among Cisco, NVIDIA, MITRE, Orion Development Company, Booz Allen, and T-Mobile, which brings together AI, compute, and wireless to create a secure, American-led path from 5G-Advanced to AI-native 6G. “I encourage Congress to lean in on areas where the United States has a strategic leadership role, such as compute, core networking, and applications,” Everson said.

Second, he called for modernizing permitting and infrastructure processes. As computing becomes more distributed, permitting must enable rapid and responsible deployment. Everson also urged the committee to consider the evolving costs of AI-ready networks when evaluating the future of the Universal Service Fund, so that rural and urban communities can share in the benefits of AI-driven connectivity.

Third, he emphasized the importance of maintaining a balanced spectrum policy. The 800 megahertz of licensed spectrum recently made available by Congress is essential to high-capacity, high-uplink connectivity. At the same time, the FCC’s authorization of the 6 gigahertz band for unlicensed use is equally important to meeting enterprise demand. “A dependable pipeline of both is foundational to American leadership,” Everson concluded, expressing gratitude for the committee’s efforts to rebuild that pipeline.

Source:Network World News