I think it’s a good time to step back to basics with my efforts in IT and open up a textbook. For this, I have selected a book; ‘Thoughtful Machine Learning with Python’ by Matthew Kirk. Using this I shall write a series of blog posts linking the central ideas to ITSM to see what comes out the other side.
Here we go 🙂
First off, the groundwork, what is machine learning and how could it relate to ITSM?
So the practice of ITSM generates a great deal of data including
- Organisations and their users
- Supported software and hardware
- Defined processes and workflows
- Fault descriptions
- Request definitions
- Troubleshooting steps and their outcome
- Solutions and fulfilment methods
- Agents completing the work
Humans make sense of this data in one way by combining it all inside a tool called the ITSM platform. In another way, they use it to perform the same operations repeatedly by different human agents. In another way, they use it to define improvements to processes or systems through problem and change management. It’s also used for reporting and evaluating a service.
So we would look to machine learning to automate some of these duties. For example through supervised learning we could build explanations of cause and effect for problem management. It could also be part of the intelligence of a robot agent that supports external clients. Unsupervised learning would extend both of these functions to define exact root cause analysis and solution delivery. Unsupervised learning would also deliver robust evaluation techniques. Reinforcement learning would support improvements to systems and processes.
The trick for introducing machine learning to ITSM is understanding the structure of the data, how it flows and grows then eventually how its quality must be improved and its consistency maintained! From that point the machine learning algorithms that provide the ITSM functionality can be developed.