Setting incident category using machine learning | Pilot program


We made an important and exciting change on Hi. Some selected users no longer have to specify a category when submitting an incident.


Why did we do this?

We received feedback that our categories can be confusing and users are not always sure which category to select. In an effort to simplify the user experience, we eliminated the step of selecting a category from the process of submitting an incident to our Customer Support department.


So how will it work?

Here is the exciting part; we do still need categories. Our internal teams use incident category to help classify and organize their work. Now, however, instead of users selecting a category manually each and every time they submit an incident, we are using some new code to assign the category automatically based on keyword mappings.


What are the details?

Our Emerging Products team has developed a product feature that we are using to model the keyword mapping. The feature is a machine learning system based on Bayes theorem. Essentially, we ran the model against all of our historical incidents to ‘teach’ our system which incident category has historically been applied to the keyword data we have. Then, the next time an incident with similar keywords is submitted, the system will know the best and most appropriate category automatically. Over time, the system learns and updates. If internal engineers adjust the category manually after the incident is submitted, the system remembers the change and uses the new information to refine the modeling. This results in a better customer experience and, hopefully, more accurate categorization with incident submission.


How do I get this for my instance?

We are still experimenting, but we hope that the model will prove flexible enough for inclusion in the ServiceNow platform in the near future.


How can I give ServiceNow feedback about this new feature?

Let us know what you think by leaving feedback on the Hi Service Portal. We hope you enjoy the simplified user experience!

Article Information

Last Updated:2016-09-26 08:06:29