I will be first to admit that marketing can sometimes get ahead of reality. It is our job to create a story that paints an appealing vision of the future, knowing that it may take a little while for engineering to actually “catch up”. And when multiple vendors start screaming the same buzzwords from the top of multiple different mountains, the line between product and PowerPoint can become a bit blurred and confusing.

“AI for IT” is a perfect example of this. Customers want an end-to-end solution for automation and insight that extends from the wireless LAN, through the wired LAN, across the WAN and even into the security domain. The aforementioned term was coined to describe this movement. For a while, there were not many vendors espousing the vision of a new AI-driven IT infrastructure, and they faced some head winds as peoples asked questions like “Is it shipping?”, “Is it really AI?”, and “How does it really benefit customers?” Below are some answers to these questions:

Yes, Virginia, AI for IT is real.  

There is a lot of debate as to whether the different techniques used by vendors can really be called “AI” or whether the claims are just being made from marketing teams gone amuck. Below are some examples from the wireless space that check multiple boxes on this front. For example:

• • Numerous data science algorithms are used to simplify operations and troubleshooting. For example, mutual information is used to classify and measure traffic for service levels, and Bayesian Inference is used for predictive analytics to determine the most likely root cause of issues. And we are leveraging the ARIMA (AutoRegressive Integrated Moving Average) time series model to detect anomalies in service levels and proactively report and drive faster resolution.  
  • Supervised machine learning correlates events for Wi-Fi troubleshooting, and unsupervised machine learning is used for calculating location using virtual Bluetooth LE  In addition, neural networks (ie deep learning) are used for anomaly detection.
  • Reinforcement Learning is leveraged by our AI Radio Resource Management in order to automatically optimize channel and power assignments by learning how these variables impact the user experience.
  • Natural Language Processing (NLP) is a key element of  virtual assistants that enable customers to engage with WLAN platforms in a manner similar to a human RF expert.  

Next steps for AI-driven IT    

Most of the industry is now espousing the notion of AI and have launched products at varying states of maturity. As a result, we have a different challenge – people now ask “How can I deploy AI effectively?”

Fortunately, there is an ecosystem building around “AI for IT” that is propelling it to mainstream. More specifically, there is an increasing number of vendors who are partnering to bring AI automation and insight across the full IT stack. Last year, for example, Mist announced partnerships with VMware (integrating with the VMware SD-WAN by VeloCloud™, Palo Alto Networks, Verizon, and Juniper Networks to deliver on this promise.  

In addition, we are already seeing customers deploy AI for IT solutions today, with Dartmouth College and the Orlando VA Medical Center being prime examples.

At ONUG Spring in Dallas, we will be taking things even further. A group of vendors and IT executives will start to move the industry’s journey toward AI-driven IT operations by presenting a framework for a common infrastructure data store. The goal of the meeting (May 8th at 9:05 am; Moderated by Nick Lippis and Mist co-founder and CTO Bob Friday) is to explore use cases/problems IT business leaders need their infrastructure vendors to solve or optimize.

These are exciting times for the IT industry as real companies are putting time, money and resources into making AI for IT happen. And it is especially exciting for me, as what started with some simple marketing slogans has turned into so much more.