As organizations accelerate AI adoption, enterprise networks are emerging as a limiting factor.
Cisco Systems’ latest report found that 75% of organizations are more confident in their AI strategy than in their network’s ability to support it, underscoring a growing gap between AI ambitions and network readiness.
AI adoption is outpacing network readiness
Titled “No Time to Wait: The Accelerating Impact of AI on Campus and Branch Networks,” the report is based on a survey of more than 3,400 IT and networking decision-makers across 15 countries.
It examines how the rapid expansion of generative, agentic, and physical AI is straining enterprise networking today.
Among its findings, Cisco reported that 75% of organizations are more confident in their AI strategy than in their network’s ability to support it, suggesting many enterprises are advancing AI initiatives faster than they are modernizing the infrastructure needed to sustain them.
Why legacy networks fall short for AI workloads
Channel Insider spoke with Michael Dickman, senior vice president and general manager of Campus Networking at Cisco, who explained that the disconnect stems from networks that were designed for traditional enterprise workloads rather than the sustained, distributed traffic patterns AI generates.
“The world has changed with AI driving business transformation, but the infrastructure underneath often hasn’t kept pace,” Dickman said.
“Most organizations have invested in AI applications and cloud services, but when they begin moving from pilots into production, they often discover their network wasn’t designed for the different and more rigorous requirements of AI traffic.”
Dickman said AI is reshaping enterprise network traffic in three key ways:
- Greater east-west traffic: AI workloads generate significantly more inter-system communication within the same site.
- More distributed workloads: AI processing is now distributed across branch offices, campuses, data centers, and multiple cloud environments.
- Sustained machine-to-machine traffic: Autonomous AI agents create sustained, high-utilization traffic that differs from the bursty, human-to-machine traffic typical of traditional enterprise workloads.
According to Dickman, these changes place fundamentally different demands on enterprise networks than traditional user traffic. As a result, organizations must address new gaps around performance, observability, and security while reducing operational complexity.
AI traffic growth creates capacity pressure
Beyond changing how data flows across enterprise networks, AI is expected to significantly increase network traffic volume.
According to the report, organizations expect AI-driven network traffic to nearly double (96%) within the first year and more than triple (209%) over the next three years. At the same time, 73% of respondents said they already face, or expect to face, capacity limitations in their network environments.
Dickman said the increase is not simply the result of more AI applications, but of new types of workloads that place different demands on enterprise infrastructure.
Transcription, video inference, and agentic AI use cases all impact the network
He pointed to several examples driving today’s AI traffic, including:
- computer vision, which relies on high-definition video and local inference
- personal productivity agents that frequently connect with both local and remote resources
- AI transcription services used by doctors, lawyers, and other professionals
“AI not only creates more traffic, as shown in the report; it also creates different traffic,” Dickman said. “Traditional enterprise networks were largely built around predictable user-to-application communications.”
He added that these AI workloads increasingly rely on longer-lived agentic sessions, sustained machine-to-machine interactions, distributed inference at the edge, and continuous automation, creating traffic patterns that differ significantly from traditional user-to-application communications.
Visibility and security challenges grow
As AI environments become more distributed, maintaining visibility has become more difficult. The research found that 71% of organizations reported growing blind spots in monitoring and visibility.
Dickman said the issue is not a lack of information but the inability to connect it across today’s more complex environments.
“The challenge isn’t a lack of data—it’s fragmented visibility,” he said.
He explained that organizations are now managing more actors across their networks, including users, IoT devices, cloud services, workloads, and autonomous AI agents, making it more difficult for IT teams to maintain visibility across their environments.
“Teams are often looking across separate networking, security, and operations tools without a shared understanding of what’s happening,” Dickman said.
To address this, Dickman said organizations should prioritize three things: shared telemetry, shared context, and AI-powered operations that can correlate information across the entire environment.
Dickman also said AI is expanding the enterprise attack surface and accelerating the speed at which attackers can exploit vulnerabilities, making traditional patching cycles insufficient. He argued organizations should embed security directly into the network through segmentation and continuous policy enforcement.
Partners see AI readiness opportunity
As organizations work to modernize their networks for AI, Dickman said the growing demand for AI readiness is creating a significant long-term opportunity for channel partners and managed service providers.
He emphasized that AI readiness represents an operational transformation rather than a “one-time infrastructure upgrade.”
“Customers are looking for trusted advisors who can help them assess AI readiness, modernize securely, implement segmentation, simplify operations, and manage increasingly distributed environments over time,” Dickman said.
“The most successful partners will be those who can bring networking, security, observability, and AI-powered operations together into a single strategy. AI is creating urgency, but customers still need trusted guidance to turn that urgency into a practical plan.”
Last month, Cisco acquired identity visibility startup WideField Security, bringing its technology into Splunk to help organizations tackle emerging AI agent security risks. Read more about the acquisition and what it means for enterprise security.