AIOps, which involves leveraging AI (Artificial Intelligence) to automate IT operations, is definitely a high growth business nowadays. Large software vendors and consultants have been moving aggressively in the sector. Then there has been the surge in venture capital that has targeted startups the category.
But none of this should be a surprise. As I pointed out in a recent post for the Atlas blog, AIOps provides a variety of major advantages. They include data-driven decision making and improved transparency; real-time monitoring and detection; predictive maintenance; remediation; and higher productivity for IT staff. But as with any technology, an organization needs to take the necessary steps to get prepared. If not, the effort could easily fail.
Then what are some of the steps to take? Let’s consider the following:
#1 – Finding the Problems To Be Solved
It’s always scary when a technology implementation is about doing something “cool” or “excitement.” This usually means that the focus is too much on the features – not the goals and outcomes.
To avoid this, it’s better to look first at the problems to be solved. For example, is the IT staff getting overwhelmed with too many alerts? Is it becoming too difficult to remediate routine issues? Is there a lack of transparency across the organization?
In other words, you should make a list of the issues. Then you can look at what realistically AIOps can solve. This will bring more discipline to the process and help to keep things on track.
#2 – Evaluate Existing Tools and Applications
Do an inventory of your IT tools. The irony is that you may already own software that provides for AIOps functions! Many vendors have been investing heavily in AI.
Next, take a look at the processes for IT. What are the workflows for incident management? Where are the bottlenecks? What needs to be changed? How is IT using existing tools? Note that AIOps is more than just selecting software. To get the true benefits, there needs to be a rethinking of the processes.
#3 – Data Plan
Data is the fuel for AI. And it can seem like magic. Data is what allows the algorithms to learn over time and get better. Yet data remains one of the biggest challenges with AI. Even companies like Google and Facebook struggle with it.
One of the reasons is that the data is usually scattered across the organization. Without getting a full view, the models will inevitably fail. It’s really as simple as “garbage in, garbage out.”
Something else: Data is always messy. There are gaps, errors, and outliers.
Because of all this, there needs to be a well-thought-out data plan. It will provide for the strategies of improving transparency and the quality. Now a good data process will be costly. But is absolutely critical for success. In fact, if there is no appetite for making the investment, then AI should not be pursued.
#4 – What Does Success Look Like?
In the early stages of an AIOps process, there is much trial-and-error. But there should still be some clear goals set. Some examples include:
- Mean time to detect (MTTD): An effective AIOps system can make significant reductions in the time it takes to identify an issue.
- Mean time to acknowledge (MTTA): This is about how best the IT team responds to the issues and the setting of the lines of responsibilities. And yes, AI can be key for automating this.
- Mean time to restore/resolve (MTTR): Note that the AI can determine if the issue can be remediated without human intervention.
#5 – Take a Gradual Approach
Modern AIOps platforms have a rich assortment of functions. Yet it is best to start with only a few.
Actually, one approach is to first use anomaly detection. This is something that will get better over time – so long as the data is good. It will also provide a sense of how to work with basic AI, which is different from traditional technologies. Consider that it involves probabilities and predictions, which can be somewhat jarring at first. Thus, there will be a learning curve for the IT team.
AIOps is definitely transformative. It is really the next-generation for IT and will provide significant improvements. But to get off to a good start, there needs to be some planning, which will not only be essential for success but also help with getting buy-in. AIOps can even be a gateway for other parts of the organization to start adopting AI. It’s a good first step in the journey.