AIOps Lessons Learned: Be Careful When Selecting a Supplier


Join today’s leading executives at the Data Summit on March 9. Register here.

The concept of AIops is simple: add artificial intelligence (AI) to IT to make operations faster and more efficient. In theory, AIops should, at best, lead to an autonomous IT environment where functions can run themselves with little or no human intervention. In practice, the path to this nirvana state is anything but simple and raises several questions. Where should you start? How do you measure the value? Is AI ready to scale in production environments? Do I need new tools?

While there is no easy button for AIops, it is well worth investigating as IT operations have become significantly more complex, especially in recent years as IT staff have been sent home to do their jobs. Businesses are becoming more dynamic, mobile, cloud-centric and distributed, creating a huge gap between budgets and the role of IT. AIops can help close that gap.

Research firm ZK Research, in collaboration with Masergy, conducted an AIops study to shed light on where the industry is located and to provide insight into how some of these questions can be answered so that companies can move forward with AIops. The study involved a total sample size of 510 enterprise-class companies across a wide range of industries.

The research mainly focused on the intersection of AIops with network operations, given the number of significant market transitions in the network and the fact that networks affect all parts of an IT system. Software-defined WANs (SD-WAN), Secure Access Service Edge (SASE), remote working and hybrid, 5G and other trends make managing the network extremely challenging using legacy operations. The aim of the study was to find out whether networking was ripe for AIops.

Here are some of the research’s key findings.

Adoption Rates Warn IT Professionals: AI Analytics Doesn’t Equal Automation

Don’t be fooled by AI-based analytics. Many IT leaders believe they are moving towards full automation because they have implemented AIops. However, that is not always true. The research results indicate that it is very possible that buyers get caught up in this misconception.

A whopping 64% of IT leaders said they are currently using AIops, compared to 37% who are merely evaluating or exploring the technology. Anecdotal research shows that the number of 64% is too high, as ZK Research expected about 50%. While this may appear to contradict the research data, it underlines that there is a misunderstanding of what AIops actually is. ZK Research said excessive adoption is often the case with emerging technology.

In the early days of the cloud, companies using a basic hosted service claimed to be using the cloud. This is a step towards the cloud, but this is not a cloud implementation. A more recent example is SD-WAN. At the beginning of the buying cycle, companies using broadband for business connectivity thought it was SD-WAN, but it isn’t.

The data represents a good news/bad news scenario for the industry. It’s certainly positive that such a high percentage of companies are interested in AIops, but it’s negative that many will go through some growing pains as IT professionals better understand which vendors are using AI and which others are using the term for marketing.

It is critical that AIops system reviewers understand what they are using and whether the AIops toolset they are evaluating can improve a user’s autonomy. AIops is not a network management tool with a few recommendations, nor is it a security dashboard with a smart color scheme or a rules-based engine that needs constant updating.

One way to find out is to look at what is powering the motor. AIops should be based on machine learning, behavioral and predictive analytics that go beyond pointing out issues and generate tickets for IT staff. AIops must provide verified solutions and help IT teams make the necessary adjustments. AIops engines must come with network management and security information, but more importantly, they must observe real-time data 24/7 to learn from existing environments.

Businesses and threat landscapes are always evolving, so AIops needs to move with them and get smarter over time. Ask the supplier about the accuracy of the product’s problem-solving ability today compared to a year ago. Other good questions to ask are: how long does it take to get to know your environment and how often do the models need to be retrained? Buyers should be diligent in selecting an AIops tool.

IT operational efficiency is a good starting point

One of the first questions asked in the survey was, “What are the main reasons for your interest in AIops tools?” as this would provide some insight into common use cases. The best comments are listed in this order:

Improve ITS operational efficiency/productivity Faster response to security risks Faster identification of security risks Improve network reliability Improve network/application performance

This data shows that companies use AIops platforms first for basic blocking and handling before using them for more advanced tasks. This is the right approach because the network must be secured and optimized before companies can look into advanced capabilities such as intent-based networking.

AIops is for network and security

One of the pleasant surprises from the survey was the convergence of network and security. Usually, large corporations keep a walled garden between the two teams. After years of talking about bringing these teams together, it’s good that companies are finally looking for a single tool that can be used by both groups.

This trend was highlighted by the question, “What are the key criteria based on which you select an AIops provider?” The best answer was “AIops features” including analytics, predictions, recommendations, and integration. This supports the above data points that the technology is being used to improve IT operations. The second highest response was “Ability to address network and security.” In fact, the survey found that 55% of those already using AIops use it in both IT domains.

Think twice before taking a DIY approach

The survey asked, “Which deployment model will you choose for AIops?” At the top of the list was in-house management at 47%, followed by co-management at 32% and fully managed services at 21%. Since it is so early in the AIops cycle, ZK Research expected a DIY preference, which is common with new technology. One caveat: Training and retraining AI models to work in a production environment is not trivial.

In this case, ZK Research recommends finding a managed service provider who will at least ensure that the datasets are good. In data science, there is an axiom that states, “Good data leads to good insights”, but bad data can lead to bad insights. Managed services can help you get the best results quickly.

AIops is no longer the stuff of science fiction. The technology is real and works today. Companies considering AIops should learn from the research and use it to solve the basics before attempting to move to more advanced capabilities. In addition, choose your supplier carefully and challenge the company to provide statistics and data on the performance of the product. Finally, if your company isn’t full of data scientists—and most aren’t—find a managed services partner who can help.

Chris Preimesberger is a former editor of eWEEK and a regular VentureBeat contributor who has been reporting and analyzing IT trends and products for more than two decades.

Zeus Kerravala is the founder and principal analyst at ZK Research. He spent 10 years at Yankee Group, previously holding a number of corporate IT positions. Kerravala is considered one of the top 10 IT analysts in the world by Apollo Research, which has evaluated 3,960 technology analysts and their individual statistics on press coverage.

VentureBeat’s mission is to be a digital city square for tech decision makers to learn about transformative business technology and transactions. Learn more

This post AIOps Lessons Learned: Be Careful When Selecting a Supplier

was original published at “”