AI is moving into more areas of business operations, and infrastructure requirements are becoming more complicated as a result. While much of the attention surrounding enterprise AI has focused on computing power and applications, the network also plays an important role in determining how effectively organizations can deploy new capabilities.
That’s because AI workloads generate substantial amounts of traffic as data moves between business locations, cloud environments, and data centers. Depending on the application, delays in moving that data can affect performance or limit where AI capabilities can be deployed. For enterprises with geographically dispersed operations, these considerations can have a significant impact on infrastructure planning.
The challenge for CIOs is anticipating how network requirements will change as AI adoption expands.
Understanding where AI workloads belong
One of the first questions organizations should address is where their AI workloads will run. Some applications may be best suited to centralized data centers or cloud environments, while others require processing closer to where data is generated. For example, a retailer using AI to analyze customer activity in stores may need to process information locally to support decisions in real time. Other applications, such as analyzing historical purchasing patterns, may be able to send data to a centralized environment for processing.
These different approaches have implications for network architecture. Rob Roache, group vice president of strategic markets at Spectrum Business®, says understanding where AI workloads will run should be one of the first considerations in infrastructure planning.
“The first priority is to understand where the workloads run and how the data is going to move. Everything else just follows from that,” Roache says.
Those decisions can affect the amount of network capacity required at individual locations, along with the reliability and performance expectations for the connections supporting them.
Planning for changing network demands
As AI becomes integrated into more business processes, organizations may discover that network capacity designed for traditional applications is no longer sufficient. Consider a retailer operating hundreds of locations. If the company introduces AI-enabled video analytics across its stores, the additional traffic could compete with existing applications for available bandwidth. Network teams would need to determine how that traffic should be prioritized and whether individual locations require additional capacity.
Planning becomes especially important when organizations expect to expand an AI application from a limited deployment to a much larger number of locations. Infrastructure limitations that are manageable during a pilot can become more difficult to address at scale.
To overcome this challenge, CIOs should also evaluate how easily their networks can accommodate changes. Adding capacity or connecting new locations may require coordination among internal teams and external providers. Understanding those dependencies early can help organizations avoid delays as AI initiatives expand.
Making connectivity part of AI planning
But there is an even more effective solution: In the age of AI, network teams should be involved in these initiatives while organizations are still evaluating potential use cases and deployment approaches. Their input can help identify connectivity constraints and determine whether existing infrastructure can support anticipated requirements.
This also gives IT leaders an opportunity to assess where greater network visibility or operational support may be needed as AI workloads become more widespread.
AI infrastructure requirements will only continue to evolve as organizations experiment with new applications and expand successful deployments. By incorporating connectivity considerations into AI planning from the beginning, CIOs can make more informed decisions about where workloads should run and how their networks will support them.
Learn how Spectrum Business® can help organizations prepare their networks for evolving AI demands.