How AI will transform the enterprise network

Self-organizing networks will no longer simply detect, diagnose and monitor, but actually take proactive, prescriptive action. Here’s how predictive AI and ML could be a game changer for IT operations and network capacity planning.


Artificial intelligence is moving beyond the realm of consumer voice commands and into the most complicated corners of the enterprise IT network.

Gartner originally dubbed the trend ‘algorithmic IT operations,’ then renamed it ‘artificial intelligence for IT operations’ (helpfully keeping the same pithy short-form ‘AIOps’, however). Forrester refers to it as ‘cognitive operations.’ Whatever you want to call it, AI is set to have a huge impact on the form and function of IT networks.

“AIOps platforms utilize big data, modern machine learning and other advanced analytics technologies to directly and indirectly enhance IT operations  (monitoring, automation and service desk) functions with proactive, personal and dynamic insight,” Gartner analyst Colin Fletcher declared in 2017.

Right now, we’re smack dab in the middle of some big-time adoption growth, according to Gartner’s forecast, which calls for AIOps market penetration to jump from five per cent in 2017 to 40 per cent in 2022. How will AI transform your enterprise network once it arrives there? We’d ask Siri or Alexa, but the answer (like most IT networks) is a tad complex.

Self-everything

AI and ML are adding a critical layer of contextual intelligence to the automation already enabled in self-organizing networks (SON). As researchers from Concordia University and the University of Waterloo asserted in an extensive paper on cognitive network management last year, “ML goes beyond learning or extracting knowledge to utilizing it and improving it with experience.”

So the more data and algorithms we put into the network, the more intelligently self-sufficient the network becomes over time.

“Networks will achieve an entirely new level of self-awareness, self-configuration, self-healing and self-protection. This will be a benefit for existing networks as well as evolved LTE, emerging IoT systems and soon to be launched 5G networks,” the authors of a Mind Commerce analyst report wrote in 2018.

This greater degree of network self-sufficiency will, of course, save enterprise organizations time, money and other resources invested in configuring, monitoring, securing and optimizing their IT networks.

Intelligent and predictive

That Canadian research team we mentioned earlier suggests AI and ML will unleash an impressive list of improved, intelligent functions within enterprise networks, including fault detection and management, performance management, configuration, accounting (such as calculating and predicting network costs and customer bills) and security (such as anomaly detection and risk mitigation).

Note that networks will no longer simply detect, diagnose or monitor these things but actually take proactive, prescriptive action on them: fixing, managing, forecasting, planning.

The predictive capabilities of AI and ML are especially promising on the security front. While common cybersecurity methods detect risk based on historic threats that are already known, the Canadian researchers pointed out that “anomaly detection using ML has also been explored to detect zero-day attacks” that have never been encountered before.

As Deloitte’s Ashish Verma told Network World, incorporating predictive AI and ML could also be a game changer for network capacity planning.

“[It] can drive new and smarter predictive insights to improve network capacity planning accuracy. This helps organizations unleash data to make more agile decisions, improve operational wisdom, avoid downtime and create a better user experience,” Verma said.

Challenges ahead

There’s always an asterisk or fine print section that comes along with any emerging tech. For AI network management, the likeliest challenges have to do with the most important ingredient in the AIOps sauce: data.

“Non-representative datasets have a severe impact on the accuracy of the model,” the Canadian researchers caution in their whitepaper. “Gaining access to representative data is not an easy task, mainly due to its sensitive and confidential nature.”

Tom Anderson, principal analyst at the Alliance for Telecommunications Industry Solutions, raised a similar red flag about data privacy in his own research note last year.

“The domain of privacy is yet unresolved,” he wrote. “This domain also intersects with public policy as these privacy concerns overlap with existing privacy regulations and may invoke new regulations directly targeted toward AI/ML applications.”

Despite concerns about privacy, Anderson concluded AI and ML carry enormous potential to “help operators improve network efficiency, lower operating costs and improve both the quality of service and customer experience.”

Early days? Yes. Yet very, very promising.

Image: miakievy/iStock

Comments are closed.