AI adoption puts pressure on campus and branch networks

 Enterprise AI adoption is placing increasing pressure on campus and branch networks, as traffic patterns become more distributed, automated and sensitive to latency.

Research from Cisco and Foundry, based on responses from 3,472 IT and networking decision-makers across 15 countries, found that only 15% of organisations have networks flexible enough to support AI at the required scale.

The findings show that 73% of respondents are already facing, or expect to face, campus and branch network capacity limitations within the next 24 months. Network traffic linked to AI workloads has increased by an average of 34% over the past year, while a further 96% increase is expected over the next year.

AI is also changing the type of traffic moving through enterprise environments. Some 67% of respondents reported increased east-west traffic, which involves communication between internal devices, systems and applications. Another 61% reported growth in continuous automated traffic generated by AI systems.

AI workloads

The pressure is expected to increase as organisations expand their use of generative, agentic and physical AI.

Respondents expect generative AI to drive an 87% increase in network traffic within 36 months. Agentic AI is expected to produce a 102% increase, while physical AI is projected to contribute a 112% increase over the same period.

Half of the expected AI-driven demand is concentrated in wireless networks within campus locations, putting additional requirements on Wi-Fi infrastructure. The research also found that 90% of respondents expect increased adoption of generative AI over the next 12 to 24 months. The corresponding figures for agentic and physical AI are 85% and 70%.

Latency can also affect whether AI systems work effectively in operational environments.

"There is a delay of about five seconds," said the Vice President of Infrastructure, Network, and End User Services, U.S.-based retail enterprise. "In those five seconds, somebody already leaves the store. So it's pointless."

Visibility gaps

Network teams are also dealing with limited visibility into AI deployments across their organisations.

AI experimentation can take place across business units without central oversight. This can make it difficult for IT teams to identify which AI services are running on the network and how they are affecting traffic.

"Right now, we don't even know what the AI-driven demand is," said the AI Strategy Leader, large U.S. technology company. "Observability is a huge gap. There is experimentation going on all over the place, and there is no way for us to really identify if somebody is deploying some kind of service on our network, whether it is a genAI solution or an agentic solution."

The research found that 76% of respondents believe their campus and branch environments require upgrades to support current and future AI workloads. At the same time, 93% said they are accelerating network modernisation initiatives in response to AI-driven demand.

Security concerns

Security complexity was identified as the most significant challenge associated with AI-driven network demand.

Respondents cited expanded attack surfaces, shadow AI activity, inconsistent policy enforcement and limited visibility into AI-driven traffic as factors affecting their ability to scale deployments.

Some 77% said AI had already expanded their attack surface over the previous 12 months. A further 78% expect security risks to increase as AI adoption expands beyond generative use cases. Meanwhile, 71% believe threats are evolving faster than existing controls can adapt, and 69% reported growing monitoring and visibility blind spots.

Although 86% of respondents said they had added security controls for AI workloads, 61% said they are holding back from further scaling AI initiatives until they have greater confidence in their security posture.

"The issue from a security standpoint is that it's hard to create the guardrails for every possible AI tool that your organization must use," explained the Vice President of Infrastructure, Network, and End User Services, U.S.-based retail enterprise.

Modernisation plans

The research indicates that organisations further along in AI adoption are taking a more active approach to network modernisation.

Among organisations with enterprise-wide AI adoption, 96% said AI has somewhat or significantly increased their network modernisation plans. This compares with 88% of organisations that have not deployed AI across the enterprise.

Mature AI adopters were also more likely to upgrade campus and branch environments to support AI capacity demands, meet compliance requirements and keep ahead of competitors. The respective figures were 55%, 53% and 51%, compared with 26%, 32% and 26% among early-stage adopters.

The research found that only 30% of aggressive AI adopters, defined as organisations with broad generative AI deployment, consider themselves fully prepared to support projected AI growth across their networks.

Business risks

Delaying network upgrades is also associated with financial and operational risks.

Three-quarters of IT leaders identified higher long-term costs from reactive upgrades or remediation as a risk of delaying modernisation. Other frequently cited risks included an inability to meet customer expectations, missed business opportunities and increased security risks.

A total of 70% cited business reputation risks, while 71% identified falling behind competitors as a potential consequence. Decreased operational efficiency was cited by 67% of respondents.

"We're just playing catch-up at the moment," said the Vice President of IT and Digital Infrastructure, U.K. education sector. "It's a worrying time, and I think it'll stay like this for another 18 months or two years."

The research was conducted by Foundry through a quantitative survey of CIOs and networking, end-user computing and technology leaders at organisations with more than 500 employees. The surveyed organisations had an average of 3,292 campus and branch locations. The research was conducted between March and April 2026.