Artificial intelligence is no longer a distant future concept; it is deeply embedded in today’s enterprise and service provider networks, fundamentally altering how data flows and how infrastructure must be designed. That was the central message delivered by Bob Everson, chief architect of provider mobility at Cisco, during a recent appearance before the U.S. Senate Subcommittee on Telecommunications and Media. The hearing, part of the Senate Committee on Commerce, Science, and Transportation, was convened to examine the intersection of AI and network infrastructure, with Everson sharing Cisco’s perspective on two critical questions: How is AI reshaping networks, and how can networks themselves harness AI’s power?

The July 30 hearing, titled “Intelligent Networks: Powering Artificial Intelligence and Transforming Communications,” brought together a diverse group of witnesses, including representatives from U.S. Telecom, Vanderbilt University, and the Nebraska Public Service Commission. Lawmakers and industry experts alike explored how the rapid adoption of AI has stressed legacy network models and what steps are needed to ensure robust, secure connectivity in an AI-driven era.

In her opening statement, U.S. Senator Deb Fischer (R-Neb), chair of the subcommittee, set the stage by emphasizing the scale of the challenge. “We will explore how widespread AI use has forced networks to evolve, requiring more capacity and more complex designs so that AI can run efficiently on those networks,” Fischer said. She also noted that private companies have already invested hundreds of billions of dollars in network deployment, with additional federal broadband programs contributing billions more for targeted deployment and maintenance.

AI changes traffic patterns and network behavior

Everson’s testimony focused on a key insight: AI is not simply increasing the volume of traffic; it is changing the fundamental behavior of networks. “AI is changing not only the volume of network traffic, but the behavior,” he said. “Cisco measured a fourfold increase in AI inference traffic over eight months. Networks have traditionally been optimized for content flowing downstream. AI 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 are active longer than conventional web transactions.”

This shift is particularly pronounced with the rise of AI agents—autonomous software that performs tasks without constant human supervision. Everson noted that in Cisco’s testing, an AI agent generated 450 percent more traffic than a person performing the same task, with roughly 70 percent of that additional traffic attributable to inference. These agents operate at software speed, creating bursts of traffic that traditional networks are not equipped to handle with conventional caching and downstream optimization techniques.

Campus and branch networks are already feeling the strain. Everson cited customer reports showing a 34 percent increase in traffic tied to AI workloads over the past year, with expectations of a 96 percent surge in the coming year. “Half of enterprise customers report that AI demand is concentrated on their Wi-Fi networks,” he said, adding that 73 percent of organizations already face or expect to face campus and branch capacity limitations within the next two years.

This is driven by significant increases in east-west traffic, latency-sensitive traffic, and continuous automated AI traffic. While much of the initial AI boom has relied on foundation models running in centralized data centers, Everson observed a growing trend toward distributed deployments. “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,” he explained. This decentralization highlights the value of the Federal Communications Commission’s 2020 decision to authorize the full 6 GHz band for unlicensed Wi-Fi use, a move that Everson credited as forward-thinking.

Key areas impacted by AI infrastructure shift

Everson’s prepared remarks laid out several key areas where AI is forcing network operators to rethink their strategies, spanning infrastructure, technical performance, cost, and data governance.

• Infrastructure: AI is accelerating the move toward edge computing. Everson argued that service providers must plan for “AI-native” traffic profiles, considering technical demands, cost implications, and growing concerns about data sovereignty and security. Rather than backhauling everything to centralized clouds, operators should push compute capacity toward the network edge—at cell sites, for example—to enable faster processing and reduce bottlenecks.
• Technical performance: Emerging physical AI applications, including robotics, autonomous vehicles, and industrial automation, may require sub-millisecond decision-making. Everson gave a vivid example: if an autonomous robot sends data to a central cloud and must wait for a round-trip response, the latency could be too high for safe, real-time operation. Localized inference and edge processing become essential to meet such demanding performance thresholds.
• Cost: AI operations generate enormous amounts of data. High-definition video analytics for public safety alone can create terabytes of data each day. Transporting that data to a centralized processing hub is not only expensive but also risks overwhelming network backhaul links. Edge processing and distributed AI architectures offer a more economical alternative, reducing congestion and enabling scalable growth.
• Data sovereignty and security: Enterprises and government agencies exhibit growing anxiety about where sensitive data resides. Many customers have security or regulatory reservations about sending proprietary or classified information across public networks to third-party cloud providers. Edge computing can help keep data within local jurisdiction, alleviating compliance burdens and strengthening security posture.

AI as a tool for network resilience

While AI workloads pose several challenges, Everson also emphasized the tremendous upside of using AI to enhance network operations. “There is a tremendous opportunity to leverage AI to deliver new applications and better performance, infuse security into the fabric of the network, and manage the increased complexity,” he said. In particular, he highlighted the role of agentic AI in changing both network traffic patterns and the administrative tools available to operators.

One such advancement is AgenticOps, which enables networks to act as self-healing systems. Everson described practical applications where Cisco’s AI-native tools automatically reroute traffic, adjust capacity, or reconfigure network nodes when performance degradation or impending hardware failure is detected. “This dramatically increases uptime and reliability for mission-critical services,” he asserted.

In addition to operational improvements, AgenticOps can help bridge the network industry’s talent gap—a persistent issue cited by many IT leaders. By automating repetitive, low-value tasks such as ticket resolution, configuration updates, and routine maintenance, network operators can lower the bar for entry-level professionals. “These tools also help close the workforce talent gap by lowering the barrier to entry and allowing more junior analysts to ramp up quickly,” Everson said. Freed from mundane chores, experienced engineers can concentrate on architectural strategy, while cybersecurity analysts can spend more time on threat hunting and detection engineering.

AI-native platforms and integrated sensing

The evolution of networks is not limited to incremental upgrades; Everson outlined a more fundamental transformation. Networks are moving from simple data pipes to AI-native platforms that function as intelligent connectivity fabrics. As compute moves toward the edge, operators can host applications directly on network infrastructure, unlocking new capabilities.

A notable example discussed by Everson is Integrated Sensing and Communication (ISAC). This technology merges wireless communications with radio-frequency sensing to detect the position and path of objects using reflections of radio waves. ISAC offers significant advantages over traditional optical sensors: it can penetrate smoke, operate in low-light conditions, and see around obstacles where conventional video analytics would fail. “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,” Everson noted.

The implications of ISAC and other edge-based innovations extend far beyond commercial efficiency. They are foundational capabilities for future smart cities, intelligent transportation, and next-generation public safety solutions. As these systems mature, they will require robust, low-latency, and secure networks that can support both communications and sensing functions simultaneously.

Policy recommendations for the committee

Everson concluded his remarks by offering three concrete suggestions for lawmakers to consider as they chart the future of national technology policy.

• Accelerate the U.S. AI-native stack. Cisco is investing heavily in AI-native networking capabilities, from 5G-Advanced innovations today to building blocks for 6G. A prominent example is AI-WIN, a collaborative venture involving Cisco, NVIDIA, MITRE, Orion Development Company, Booz Allen, and T-Mobile. The project marries AI, compute, and wireless technology to create a secure, American-led path from 5G-Advanced to AI-native 6G. Everson encouraged Congress to “lean in” on strategic areas such as compute, core networking, and applications where the United States has a unique edge.
• Modernize permitting and infrastructure rules. As computing becomes more distributed, outdated permitting processes could slow network deployment. Everson urged the committee to address these hurdles while also reconsidering the Universal Service Fund to reflect the evolving costs of AI-ready networks. His concern is that rural and disadvantaged communities might otherwise miss out on the benefits of advanced AI connectivity.
• Maintain a balanced spectrum policy. Everson emphasized the need for both licensed and unlicensed spectrum. The 800 megahertz of licensed spectrum recently made available by Congress is crucial for high-capacity, uplink-intensive applications, while the FCC’s authorization of the 6 GHz band for unlicensed use is equally important to meet enterprise demand. He insisted that “a dependable pipeline of both is foundational to American leadership” and thanked the committee for its ongoing efforts to rebuild that pipeline.

Everson’s testimony underscores a pivotal moment for the networking industry. As AI continues to infuse every layer of the technology stack, the systems that carry data must become smarter, faster, and more flexible than ever before. The debate in Washington—and the decisions that follow—will help determine whether the nation can build the infrastructure needed to sustain AI leadership, strengthen economic competitiveness, and ensure equitable access to the digital future.

Source: Network World News