Matilda Bailey is a distinguished networking specialist whose work defines the cutting edge of cellular and next-gen wireless solutions. As the industry grapples with the transition from pilot AI programs to full-scale enterprise deployment, her insights into how these technologies strain existing infrastructure are more critical than ever. She has spent her career navigating the evolution of connectivity, and her current focus on the “networking supercycle” highlights a pivotal moment where campus and branch networks must undergo a radical transformation to survive the AI era.
The discussion explores the explosive growth of AI-driven network traffic, which is projected to triple for major adopters, and the shifting patterns of internal communication. We delve into the critical observability gaps that leave IT departments struggling to identify what is running on their wires, as well as the security barriers that have led a majority of organizations to pause their AI ambitions. Finally, we examine the rise of autonomous AI agents and why traditional network designs, once optimized for simple cloud access, are no longer sufficient for the unpredictable demands of a modernized enterprise.
The data shows a staggering 209% projected climb in traffic over the next three years—how is this shift fundamentally changing the way we look at campus and branch network architecture?
We are witnessing the dawn of what is being called a “networking supercycle,” where the infrastructure is finally catching up to the raw processing power of AI. Over the last 12 months alone, organizations have already reported a 34% increase in AI-related traffic at the branch level, and for those deploying AI broadly, total network traffic is expected to triple. This isn’t just about adding more bandwidth; it’s about a total architectural redesign because 73% of IT leaders expect to hit severe capacity constraints within just two years. We are moving away from the era of predictable SaaS and CRM traffic toward a landscape where the network must be as dynamic as the models it supports, or else the entire system will grind to a halt under the weight of its own data.
While much of the AI conversation has focused on massive data centers, why are campus and branch environments suddenly becoming the new frontline for performance issues?
For the past two years, IT leaders have been obsessed with GPUs and cloud platforms, but they are now realizing that the “last mile” of the enterprise—the campus and branch—is where the real pressure is mounting. As organizations move beyond generative AI pilots and start deploying autonomous agents, the traffic patterns are shifting from simple external requests to complex internal dialogues. In fact, 67% of respondents noted that AI workloads are significantly increasing east-west traffic between internal systems and applications. If the network connecting employees and devices isn’t modernized, all that expensive data center power becomes inaccessible, effectively creating a high-performance engine with a clogged fuel line.
There is a growing concern regarding “observability gaps” in modern networks; how does the lack of visibility into AI-driven demand complicate day-to-day operations?
The current state of observability is a massive hurdle because many IT organizations literally do not know what is running on their own networks. With employees and different business units experimenting with various AI tools, there is a chaotic “shadow AI” effect where services are deployed without centralized oversight. This makes it nearly impossible to identify if a spike in traffic is a legitimate agentic solution or a rogue process, leaving teams to guess at capacity needs. Without the tools to distinguish these unpredictable patterns, 93% of IT decision-makers feel forced to accelerate their modernization efforts just to regain a basic sense of control and visibility over their environments.
With 80% of leaders stating that AI has expanded their attack surface, what specific security hurdles are causing companies to delay their deployments?
Security has become a major bottleneck, with 61% of organizations actively delaying additional AI deployments until they feel more confident in their security posture. The difficulty lies in the fact that it is nearly impossible to create static guardrails for every possible AI tool that a global organization might need to use. As AI agents begin to communicate continuously with other systems, they create new entry points and vulnerabilities that traditional firewalls weren’t designed to handle. This expansion of the attack surface is a primary driver behind the urgent push for new security frameworks that can adapt to the fluid nature of AI workloads.
As we look toward the next two years, 85% of organizations expect a surge in the deployment of autonomous AI agents. How do these agents differ from traditional software in terms of network demand?
Traditional software typically follows a predictable path, like a user accessing a cloud-based CRM, but AI agents are entirely different animals that create unpredictable, high-volume traffic patterns. Instead of a single request-response cycle, you might have three different AI agents talking to each other simultaneously to solve a single complex problem, which dramatically spikes internal east-west traffic. This shift is why only 30% of aggressive AI adopters say they are fully prepared for the projected growth; the sheer unpredictability of agentic communication defies the “steady-state” logic that networks have relied on for decades. We are moving into a world where the network doesn’t just transport data—it must intelligently orchestrate the constant, heavy-duty interactions of autonomous digital entities.
What is your forecast for the future of enterprise networking as AI transitions from a pilot phase to a core operational necessity?
My forecast is that we will see a massive, non-negotiable wave of infrastructure “ripping and replacing” over the next 36 months as the 93% of leaders currently modernizing their networks hit their stride. The industry will move away from static, centralized architectures toward highly distributed, AI-native networks that can self-heal and reconfigure in real-time to manage the tripling of traffic volumes. Security and observability will no longer be “add-on” features but will be baked into the silicon of the networking hardware itself to manage the 80% increase in attack surfaces we are seeing. Ultimately, the successful enterprise of 2027 won’t just be the one with the best AI models, but the one with the most resilient and transparent network capable of fueling those models without bottlenecks.
