Machine learning for IT Support Services:
One of the critical areas for IT services has always been faster resolution of incidents and detecting problems in infrastructure even before they occur.
With Machine Learning models in ITSM tools we can improve the customer experience, address the user issues more efficiently and in some cases without an intervention of a service desk agent. Well, this is not about automating a job and saving some bucks but, working efficiently in the system and utilizing the same resources on more important tasks. (Yes, it does save your money at the end as a byproduct. :) )
This is not just a theory since there are already a few tools which claim to use these methods.
Auto suggestions and recommendations:
To look at it from a service desk perspective, agents can achieve more FCR(First call resolution) by focusing on user’s issue and not on logging an incident. Through classification algorithms, we can train the machine to classify incidents and automate the incident logging process.
Now tools can intelligently categorize, prioritize and assign a ticket just when the agent puts in the description. An example would be;
User reports the issue via call or email. Agent can simply put the issue description in the ITSM tool and it automatically categorizes, prioritizes and logs the incident. The tool can, not only suggest a full description while you are typing but come up with other intelligent solution using Knowledge Based articles which could resolve the issue. This in return either increases FCR per agents or allows them to reach out to correct responsible team.
An another illustrative example would be, user sends across an email or logs an incident/SR on service portal for password reset, the ML algorithm, based on its training from past similar data, can identify the user and reset the password automatically. This single issue accounts for around 20% of the SRs in most of the organizations.
To proactively detect problem in the infrastructure, the classifier model could derive insights from incident data and suggest which area is painful for the infrastructure and the business.
I have seen organization using resources on analyzing server generated alerts and deciding whether this needs anyone’s attention.
Machine learning models can help us analyze these alerts and incidents and suggest an optimum solution for it. For example, if it finds that there are multiple alerts logged for max memory utilization, it will log a CR to increase the memory. (Just an example)
Many would argue on this since we already have ITSM tools which monitor server performance and logs incidents automatically. However, this just brings an engineer’s attention to it and all he/she does is temporary resolving the issue. It does happen sometimes that a smart engineer brings this up to the management for memory upgradation, but again, this consumes time.
The important thing to understand here is, it’s all about data which the ITSM tool is flooded with.
These are just two possibilities that I have brought attention to. In fact Machine Learning can improve the ITSM workflow from predictive analytics and maintenance to demand planning and many more.