HaloPSA Integration & AI Classifier is now LIVE!

Welcome to tribu

Tribu Triage Classifier

First Release

This capability has been under beta testing and is now available to all Tribu partners. This first release has limited customisations and is an introduction to machine-driven predictions for your helpdesk service. Because of this, we are seeking your feedback on ways to improve and implement it for day-to-day use. Feel free to provide feedback here.

About the Tribu Triage Classifier

The Tribu Triage Classifier is an NLP-driven classifier that returns two text labels. An issue type and a sub-issue type. The model used in this classifier has been trained from several years of IT Support tickets from various backgrounds of IT Support helpdesk disciplines. The labels produced are a trial set of labels with varying levels of granularity. This classifier takes in the description text of a support ticket to produce the two labels. At this time, the model testing results yielded 92% precision and 91.5% recall for an F1 result of 91.52%.

Use Case

The Tribu Triage Classifier is intended to provide the initial Issue and Sub-Issue type labels for IT Support tickets as they arrive at your IT Support helpdesk. By implementing this classifier, you can make use of predicted classifications to automatically trigger any given workflow you have set up, or set up new workflows to take advantage of these predictions. By doing so, it is intended that the workload of first-response evaluation, ticket routing, and next steps can be removed from the human resources of a helpdesk, allowing those team members to do more meaningful, engaging, and hopefully, enjoyable workloads.

Setting up the Triage Classifier

The Classifier will insert labels into two custom fields in your ticket management system. Choose the PSA/Ticket management system from the below:

Datto Autotask

ConnectWise Manage

We do not currently update the default Issue and Sub-Issue type fields that exist in Autotask or Manage as the labels the classifier uses are custom and will not match your own labels.


In this article