AI in ITSM equals ‘service intelligence’ — ITSM’s next frontier

The first and most important aspect to understand about service intelligence is that it applies standardized machine learning and AI algorithms to ITSM data. In order to fully understand (or devise) in which areas service intelligence can help organizations, some basics on AI algorithms are required. Although there are hundreds of potential applications, in this article I will focus on some examples that are specifically used in the context of using AI in ITSM.

1. SLA performance prediction

Using correlation and regression models applied to the assets with which services are delivered, it is possible to build predictive models that indicate the likelihood that an organization will be able to achieve its service-level agreement (SLA) and performance indicators. With these insights, organizations are able to adjust focus in time so that they can allocate resources more efficiently to optimize SLA performance.

2. Preventive CI maintenance

By grouping and analyzing incidents to configurations items (CI), it becomes possible to see which elements (or infrastructure components) are most likely to cause disruptions to service delivery. With this information, organizations can work to make their services more resilient, or replace components that are likely to cause problems toward the future.

3. Text analytics support routing

Using text analytics algorithms and analysis tools, incidents can be categorized into groups and, subsequently, automatically assigned to the person (or team) that is most skilled to deal with the specific request. Especially in larger organizations with different specializations, this can greatly enhance speed. Request and problems are immediately sent to the right people, and it is possible for every support analyst to specialize into different areas.

4. AI-driven phone support

Using sentiment analysis algorithms, which can detect emotions in the words or speech of a customer, systems can learn which services are appreciated by customers and which can be further improved. With these solutions, it is possible to customize the response toward customers based on the emotions they are experiencing. A different answer is given when a person is detected as being angry, instead of when that person is happy.