AIOps Evolution Podcast | Season 2
Interview with Leslie Minnix-Wolf from ScienceLogic
Many IT leaders would agree that AIOps is no longer a “nice to have” strategy, but a necessity as data points balloon in their IT environment. Today’s IT teams cannot keep up with the complexity, and AIOps is a viable strategy to get ahead of the curve.
The larger question that we’re tackling in this AIOps Evolution Podcast is which AIOps use case to start with? In this episode, Sean interviewed Leslie Minnix-Wolf, VP of Product Marketing at ScienceLogic on the question that many IT leaders ask: Where do I start my AIOps Journey?
TL; DR
AIOps Shifts the Load
What is catalyzing the AIOps movement right now? In a few words, the noise volume and cloud migration. It’s a nascent market formed out of the enterprise monitoring market. Organizations are dealing with more information and more technology, which produces more opportunities for things to go wrong.
As the IT environment grows exponentially complex, professionals and IT leaders need efficient methods for understanding what’s happening. People are looking for a new approach that is both proactive and predictive. AIOps shifts the load off of data personnel by diagnosing and resolving issues, sometimes without human intervention.
As IT leaders look for ways to integrate an AIOps strategy into their organization, the main question is where to start the AIOps journey and what use cases will produce the most ROI?
How to start the AIOps Journey?
There are so many use cases for AIOps, but knowing where to implement it—especially if you are starting from the ground up–can be daunting. Leslie says, “It’s about the data.” So start there.
A study recently revealed that data scientists spend about 80% of their time on collecting, cleaning, and preparing data. That’s time that could be spent on resolving issues and optimizing the IT environment. If an AI can shift that load of data processing from the data scientist, that optimizes productivity and resource utilization.
As IT leaders consider AI for IT Operations, they need to ask what tools and mechanisms are in place? Is there a data lake that is cleaned and prepped to drive automation? Before considering automation, IT leaders must have data that will provide the AI with accurate information. Cleaning up data is a powerful AIOps use case.
What use case should you select?
AIOps Use Case #1: Understand the problems in your environment
As with any organizational change, it’s best to start small with a use case that will have some meaningful impact. This is new so you are likely going to have some failures. By starting small you can limit your early exposure and get your feet under you. With AIOps, data is the root of all knowledge and drives processes.
Once you have your data together, you have to understand what problems exist in the environment and what problems are worth investigating.
Additionally, you will make a much better pitch to your leadership team if you understand the business or service impact of those problems. A successful AIOps use case serves to better the organization.
AIOps Use Case #2: Diagnose the problem
Once an IT team has an issue identified, the next step is figuring out what caused it. Leslie noted that manual root cause analysis and diagnosis of issues can take upwards of 90 minutes or more depending on the criticality of the issue and the data available. For a business, every minute of downtime means loss. According to Gartner, the cost of downtime is about $5,600 per minute. Multiply that by 90 minutes. That’s a downtime of over half a million dollars.
Root cause analysis is an excellent AIOps use case to step into after integrating AI into your data lake. It pinpoints the root cause of an error in real-time and allows teams to focus on repair and resolution, significantly cutting mean time to repair (MTTR). A faster resolution to downtime means keeping operations running; in business terms, that means revenue protection.
AIOps Use Case #3: Automate repair
The next logical use case would be automating repair. What if it was possible to not only use AI to troubleshoot, but to repair the issue as well? Here’s where to start with this third AIOps use case.
Leslie emphasizes it’s not about fixing problems, but about leveraging the data that you’ve collected. An example would be feeding data into a configuration management database, automating the ticketing process, or integrating with change management processes.
As people move towards the DevOps approach, it’s about being able to feed the data that you’re monitoring into the DevOps processes and the DevOps data into the monitoring processes from an operational perspective. That way you better understand when changes are being deployed (e.g. new release, new app, new service).
Changes are the most common cause of a problem; so being able to feed data into the daily monitoring of operations is crucial.
Once you’ve got your data and it’s usable, being able to automate workflows around DevOps, SecOps, ITSM processes, whether it’s change or service management, constitutes a strong use case for AIOps.
Where is the ROI in an AIOps use case?
What should people be thinking about in terms of ROI and investment value from AIOps? IT leaders have to think about AIOps in terms of how it can benefit the business. Some key value metrics include the following:
- Bringing data together
- Saving time
- Leveraging tools you already have
- Having a single platform that houses data
- Standardizing data
From those metrics, we tend to see the ROI come from tools elimination, the infrastructure they run on, the admin staff, the training, the integrations organizations build. Research shows that automation with AIOps produces significant ROI. Leslie pointed out several use cases in her own experience to back this up.
ScienceLogic Case Study: Saving Time Saves Cost
On average it takes customers 30 minutes to key in information to the service desk. Manually populating a ticket takes time and resources. Automating that task generates tickets faster and enables people to begin working on resolving the issues sooner. The result is that people are fixing problems as opposed to opening tickets.
Leslie pointed out another use case from customers in routing tickets. It takes around 90 minutes to route a ticket to the right person. Most companies struggle to route the ticket because they don’t have good data. The definitions of their assets are unclear (e.g. who has the asset, who owns the asset, who is responsible for the asset).
ScienceLogic had a customer who had 200,000 tickets in a year. Only 2000 of those tickets were routed correctly, because they were not automatically repopulating the CMDB with updated data. They felt they were chasing around a problem. The solution was to automate the data in the service management desk database. This had several advantages:
- Reduced the time it took to create tickets
- Reduced rerouting
From these use cases, we are able to quantify the value of AIOps just on the time and resources it saves on ticketing alone.
Huge Gain$ with Mature AIOps Use Cases
It really starts to pan out with more mature customers who have automated the collection of diagnostics and the triaging process. When a problem occurs, AIOps collects information immediately at the point of error. It sends that information to the operations team with a very clear message stating what’s important that is causing the service to degrade.
Leslie presented another ScienceLogic Customer, Cisco, that saves ~$14 million from automatically enriching the events with diagnostic information when there is a problem with Cisco equipment. That information enriches their services to their clients. Saved time and staff productivity converts to cost savings that allows them to eliminate steps that an admin or operator would have to perform.
AIOps is not a leap, it’s a step up
The final goal of any AIOps journey is full integration and implementation with the organization’s infrastructure, people, processes, and technology. How do you know if you’ve successfully integrated the AIOps use case? One key indicator is your IT team’s acceptance and confidence in the system’s delivery. The other, and most important indicator, is demonstrable ROI in either cost savings or an impact on revenue.
The more the machine can learn from your data, the more it can recognize patterns. If it knows what is happening on a recurring basis, it can analyze the data collection and remediate with machine learning. People will have confidence in the system due to these past successes that have proven to them that the machine is doing the right thing to resolve the issue.
The value of AIOps only increases with the more use cases implemented. However, it doesn’t have to happen overnight – and it won’t. AIOps is a journey, and you don’t have to jump in with two feet on the first day. As many AIOps use cases have proven, most IT teams start with something small and still see tangible results that produce ROI and value.
