Enterprises are investing heavily in AI tools and projects, but network upgrades are all too often left off the investment list. Today's networks were built for rapid transaction throughput, emails, web browsing, file transfers and database queries — but can they handle daily AI on corporate and edge shadow networks, and what should CIOs and network managers do now?
With scant empirical knowledge, no one really knows how AI will affect IT or networks. From a network perspective, AI will demand bandwidth and throughput for large data payloads, and with departments throughout the company likely to use AI in on-demand mode, it will be difficult for network staff to preconfigure network bandwidth and runtime priorities based on a regular and predictable schedule.
This is why it's essential to catalog and evaluate the network's present capabilities. How is the network performing when it comes to bandwidth, latency, security, reliability, quality of service and network management? Are there present performance bottlenecks? Are routers, servers and access points up to date?
Related:Mastering the architecture of hybrid edge environments
Performing a network assessment gives organizations a baseline from which they can plan their network upgrades for AI.
It's practically a given that almost every department within an organization will use some form of generative AI, even if it's only a commercially available version of ChatGPT. However, there will be areas in the company that will want to go far beyond that. They will want to purchase or build AI systems that can predict supply chain performance, produce medical diagnoses or assess financial risk.
Historically, business analysts and IT developers met with users during these early project conceptual phases, but given the vital role that the network will play in delivering AI, this is no time for network managers to sit on the sidelines.
Instead, network managers should insist on being part of the early conceptual and planning stages for AI systems, as these projects will likely require network upgrades. Network managers should also develop ways to explain the technical requirements of the network in plain English to other decision-makers, so that everyone is on board with any upfront network investments needed to support AI.
At the end of 2025, 5G technology was closing in on 100% coverage in North America, but some companies still lagged behind with routers and switches that were incompatible with 5G.
Related:Will 2026 be the year of data center restructuring?
If networks are to be readied for AI, the routers and switches they use, minimally, must support 5G communications and, ideally, be upgradeable to 6G when it becomes available.
The same goes for software that runs and manages the network. If a company's existing vendors lack roadmaps for scaling their products for AI, new and scalable products should be considered. Installed bases of servers, as well as SAN and NAS storage, should also be upgraded for AI.
In all cases, the goal for network managers is to request AI-ready technology upgrades that will have at least a five-year lifespan, since no one wants to go back to the C-suite in a year or two to ask for another upgrade.
One way around the scalability conundrum is to deploy AI in the cloud, where you can incrementally scale your network resources as you need to. This approach makes sense for two reasons:
There already is a history of companies scaling up their cloud resources and spending that management seems to accept.
Since no one really knows what type of resources the company's AI will require, you can scale up or down in the cloud without the risk of acquiring network assets that you may not need.
Cloud-based AI can be deployed as a private cloud for sites that want airtight protection and security controls for their AI. Companies also have the option to share resources and spend in the cloud with others. Similarly, choices can be scaled for cloud AI support. These choices range from the network staff choosing to support their own AI infrastructure in the cloud to staff opting to have the cloud provider do this.
Related:It's time to revamp IT security to deal with AI
Companies using AI in multiple departments should consider implementing zero-trust networks if they don't already have them. A zero-trust network will not grant network access to a user if the user lacks the proper access and permission credentials. Zero trust also includes built-in auditing capabilities that automatically alert network staff if it detects any IT assets being added to, removed from or modified on the network.
Zero-trust networks go hand in hand with strong user identity and access management for users moving between on-premises and multi-cloud data with AI. There are three separate identity management technologies that network managers should consider:
Network managers want to be proactive with AI by providing scalable, pliable and optimized networks that can carry the AI load. At the same time, they don't have much experience with AI and how it will impact networks.
By inserting themselves early into AI discussions and projects, network managers can better prepare themselves to ensure that the network expected to process and transport the AI is up to the task.
President of Transworld Data
Mary E. Shacklett is an internationally recognized technology commentator and President of Transworld Data, a marketing and technology services firm. Prior to founding her own company, she was Vice President of Product Research and Software Development for Summit Information Systems, a computer software company; and Vice President of Strategic Planning and Technology at FSI International, a multinational manufacturer in the semiconductor industry.
Mary has business experience in Europe, Japan, and the Pacific Rim. She has a BS degree from the University of Wisconsin and an MA from the University of Southern California, where she taught for several years. She is listed in Who's Who Worldwide and in Who's Who in the Computer Industry.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。