AI is moving from experimentation into everyday enterprise use. Employees are using AI assistants to work with information, customer applications are becoming more intelligent, and cameras, sensors and other connected systems are generating data continuously.
Most discussions around these deployments focus on models, computing power and data. One question stays in the background: is the network connecting it all ready? By Cisco’s own estimate, only 15% of organisations currently have a network flexible enough to support AI at Scale.
Even the most capable application can stumble if data can’t move fast enough, from the device that creates it, to the system that processes it, to the person who acts on it. For CIOs, this makes network readiness a core part of AI planning, especially as pilots give way to real deployments across physical, operational environments.
AI Is Changing the Pattern of Network Demand
Traditional enterprise applications tend to have relatively established traffic patterns. AI-enabled applications can create more varied interactions between endpoints, applications and computing environments.
Consider an organisation using computer vision across a facility. Cameras may continuously generate information, while the application analyses selected data and sends relevant outputs to an operations team. The requirement is different from simply giving employees access to an online application.
The question is no longer only how much traffic a network can carry. CIOs also need to understand where traffic originates, where processing happens and whether particular applications require consistently fast responses.
This calls for closer alignment between application architecture and network planning.
The Indoor Environment Is Becoming Part of the Technology Stack
Large buildings present a separate challenge because several forms of technology operate within the same physical environment.
An office, hospital or airport runs employee devices, visitor Wi-Fi, security systems, building management and connected sensors, all at once, part of a global device base projected to cross 29 billion connections by 2030. AI adds more: video analytics, intelligent access, automated monitoring.
These systems do not necessarily place the same demands on connectivity. Some require high capacity, while others depend more heavily on consistent coverage or predictable performance in specific locations.
This makes the physical layout of a building relevant to technology planning. Network design has to account for where devices are located, how people use different spaces and how demand changes throughout the day.
Edge Computing Changes Where the Network Has to Reach
The expansion of edge computing adds another dimension. Some AI workloads can be processed close to the source rather than sending all information to a central cloud environment. This can be useful where large volumes of data are generated locally or where an application needs a rapid response. The shift shows up in the numbers: the global edge computing market is on track to grow from roughly $65 billion in 2026 to over $270 billion by 2030.
But distributed processing creates additional points that need to be communicated. Devices may connect to an edge location, which in turn needs to exchange information with enterprise systems and cloud platforms.
For organisations operating across several buildings or sites, this can create a more distributed infrastructure environment. The network consequently has to support different computing locations rather than assuming that every workload begins and ends in a central data centre.
Resilience Cannot Be an Afterthought
As Al embeds into operations, infrastructure failures carry heavier consequences. Enterprise networks already run ERP, point-of-sale, cloud applications, communications and security. Add Al to these workflows, and a single connectivity failure can ripple far beyond the application itself. Recent industry research puts the median cost of an enterprise outage at close to $9,000 a minute.
Resilience therefore needs to be designed into the architecture. Alternative routes can keep traffic moving when a primary connection fails, while automated re-routing can reduce the need for manual intervention. Monitoring can identify emerging problems, and recovery arrangements need to be tested before they are required.
The objective is not to eliminate every possible failure. It is to prevent one failed component from becoming a wider business disruption.
One Infrastructure Layer Can Serve Multiple Needs
There is also a case for reconsidering how connectivity is deployed across large environments. Building separate infrastructure every time a new application arrives can increase complexity and result in assets being used inefficiently. A shared infrastructure approach can allow different applications and users to operate over a common underlying layer while maintaining the performance requirements of individual services.
This is particularly relevant in high-density environments where connectivity is required for several purposes at once.
At iBUS, we see this in environments where employee access, visitor connectivity, building systems and other digital applications operate within the same physical space. Planning the underlying infrastructure around these overlapping requirements can provide greater flexibility as new technologies are introduced.
Planning for Applications That Have Not Arrived Yet
The difficulty for CIOs is that today’s AI deployment may not resemble tomorrow’s. An organisation could begin with employee productivity tools and later introduce AI-enabled customer services, connected operations or applications using edge processing. Designing infrastructure around a single use case can therefore create limitations later.
The more practical approach is to establish a clear understanding of the enterprise environment: the locations involved, the devices being connected, the applications being supported and the routes through which information needs to travel.
This does not mean building excess capacity without purpose. It means making infrastructure decisions with enough flexibility to accommodate a reasonable range of future requirements.
The Infrastructure Opportunity
Al is putting a spotlight on parts of enterprise architecture that have long stayed in the background. For CIOs, that is an opportunity, a chance to ask whether infrastructure decisions are keeping pace with deployment. Cisco’s AI Readiness Index found that 83% of organisations plan to deploy autonomous Al agents, yet only one in three believe their infrastructure is actually ready.
The answer will not be identical for every organisation. A hospital, manufacturing facility and corporate office will have different requirements. What they share is a growing dependence on the infrastructure connecting their digital systems to the physical environments in which they operate.
That is ultimately an engineering question: how do you build an environment that continues to work as the demands placed on it change?
As Engineers’ Day highlights the role of engineering in designing systems for real-world conditions, the same principle applies to enterprise infrastructure. The strongest architecture is not necessarily the one built for today’s peak requirement. It is the one designed thoughtfully enough to accommodate what the enterprise is likely to ask of it next.
Authored by By Ram Sellaratnam, Group CEO & MD, iBUS Network and Infrastructure