Today’s IT networks are complex and at times unwieldy.

Views of network traffic can be blurry, it’s often difficult to determine the cause of an outage, bad actors are trying to get in all the time — those are just a few things network operators are dealing with everyday.

Some experts say this could all change — or at least dramatically improve — with AI, specifically AI networking, which as the name implies, merges AI and networking infrastructure to automate and enhance IT operations.

AI networking is still an emerging concept, and many use cases are as yet aspirational, but some say it has the potential to completely transform networking.

“It gives us a much more realistic opportunity to aggregate data, coalesce it, do analytics on it in real time,” Forrester principal analyst Carlos Casanova told SDxCentral. “It’s moving more into this proactive and predictive type of scenario, getting ahead of these larger events.”

AI networking could ‘disrupt’ traditional networking operations

AI networking is a subset of AI for operations (AIOps) that “offers great potential to disrupt long-standing traditional networking operations to create a massive productivity increase,” according to Gartner (which coined the term in 2023).

Notably, AI networking addresses so-called “day 2” functions around management, maintenance and optimization. It can help boost network performance, while also supporting security and incident management, improving IT service management (ITSM) and increasingly delivering day 0 (conception) and day 1 (deployment) capabilities.

Essentially, “it’s infusing intelligence into the networking fabric,” Will Townsend, VP and principal analyst at Moor Insights and Strategy, told SDxCentral.

AI networking leverages a multitude of tools, including generative artificial intelligence (genAI), natural language processing (NLP), predictive analytics, trend analysis, pattern recognition and event correlation. In more advanced scenarios, it can also incorporate digital twins.

While AI networking is still in its early stages, Gartner estimates that by 2027, 90% of enterprises will be using AI to automate day 2 network operations. That’s compared with less than 10% in 2023.

“There has been much hype around [this technology] prior to late 2022 when ChatGPT brought generative AI into the forefront,” Gartner analysts write. The OpenAI chatbot has “driven hype to a new level” because genAI is demonstrating real capabilities, such as network configurations and troubleshooting.

From identifying bottlenecks to predictive maintenance

AI networking can perform numerous functions, including:

• Automating network configuration, monitoring and troubleshooting, incident management, software updating and recommendation and response.
• Optimizing ITSM. NLP can field the simplest, most common service desk inquiries. AI can also reduce false-positives before they reach the desk, and can hand off higher-level issues to humans.
• Performing predictive maintenance through study of patterns, anomalies and trends.
• Identifying bottlenecks and congestion areas. In more advanced environments, AI can allocate resources, scale infrastructure and reroute traffic.
• Determining probability of hardware failure
• Identifying anomalies that could be the result of security breaches or attacks.

“There’s a lot you can do in a semi or fully automated manner,” said Forrester’s Casanova.

For instance, AI can help with simple tasks in the continuous integration/continuous delivery (CI/CD) pipeline, he noted. If code is automated, developers can go through check-ins and check-outs in code libraries when a system is “going off the rails” and experiencing degraded performance. Upon quick analysis, software engineers can determine that a piece of code has diverged from the code library — then replace it, spin up another instance or redirect to another cluster.

Similarly, AI can help network operators identify what caused network failure (and where and why) and redirect network traffic to a different data center or cloud infrastructure when latency issues occur.

“It’s providing more consistency from a performance standpoint,” said Townsend.

When it comes to ITSM, AI could be “incredibly enriched” with past ticketing information, Casanova pointed out. If a certain identical incident or question keeps coming in, AI can handle it and follow past action steps that resolved the problem.

Suggestive alerting provides an upskilling opportunity, too. Tier 2 work (which provides more specialized support) can be moved to the Tier 1 helpdesk to provide new work experiences, said Casanova.

‘AI in the loop’ everywhere

Organizations have been trying to nail down configurations management — that is, the engineering process that tracks and controls IT services and tools across an organization — for decades, Casanova pointed out. They continue to struggle with a multitude of issues, including flexibility, visibility, agility, security, deployment, operational knowledge, making and measuring regular changes, validating those changes, and others.

The goal is to eventually get to a point where organizations can make decisions based on real-time network information “to provide actionable insights to service desk folks, engineers, operations folks,” said Casanova.

Ideally, AI will help network users identify issues before they occur in a “proactive and predictive manner.”

Look at it as “AI in the loop,” said Casanova. “Everything we do is going to have an infusion of AI.” The combination of humans and technology will tremendously support operational practices and “enhance human judgment.”

With AI, ideas don’t have to be as aspirational, he and others point out. “Technologies are coming together now in more realistic opportunities,” said Casanova. “If you apply them properly, you have a much greater opportunity to achieve some of these greater visions that we’ve had for 20 years.”

Going forward, said Townsend, genAI’s role in supporting AIOps and AI networking has a potential “factor of 10 or 20 or 30X what we’ve seen with the cloud.”

With genAI and natural language interfaces, network operators can also get closer to the concept of intent-based networking, which has more predictive capabilities.

Further, “the whole notion of a self-healing network is a topic that has been talked about for years,” said Townsend. “I do think it’s realistic with where we’re headed.”

Modern enterprises must adopt AI (and learn to trust it)

But AI networking is still a relatively young concept, and there are challenges in implementation. For instance, the mixing of phrasing — AIOps, AI networking, observability — has some confused, and enterprise leaders are seeking more clarification on how it all works.

Many leaders, even as they know how transformative AI can be, have reluctance around overall adoption, said Brandon Butler, research manager with IDC’s network infrastructure group. Organizations simply need to use AI systems more regularly to become more comfortable with their functionalities, he said.

Start out with low-hanging fruit capabilities, Butler advised, such as having AI analyze network traffic, flag degradation or security events, perform root cause analysis and provide recommendation and guidance.

“There’s a general apprehension amongst organizations to hand over the keys to a fully self-driving network,” said Butler. “Will we get there one day? Possibly.”

Casanova agreed that organizations should use AI “to the point that they’re comfortable with.”

“It is definitely not a one size fits all by any stretch,” he said. “How you go about it versus how I go about it are going to be slightly different.”

Other concerns revolve around lack of quality data to properly use AI, inaccurate recommendations causing even bigger problems, technical debt and growing requirements for new skills — not to mention inflated expectations and the struggle (in some cases) to get cultural buy-in.

However, hesitation is more dangerous than anything, Casanova emphasized. If organizations want to be high-performing and keep pace with change, they must embrace AI.

“The networks around us, they’re moving, our peers, they’re moving,” he said. “We have to keep moving.”