6 October, 2025

AutoQA

We've done a couple of updates on the system prompts to:

• Prevent AI from mentioning explicit guidelines which are ment for internal use.

• Enforce AI to share what a correct solution or approach would be, if one is mentioned in the instructions or the linked KB articles.

In addition, we've fixed a few bugs with its input and output:

• Fixed the input parsing to detect URLs in HTML messages.

• Enforce AI to provided a comment and root causes for markdowns.

• Do not provide root causes for full scores.

• Do not cite criteria or other things which are not KB articles.

Criteria

We have improve the AI generation of root caues to make it more accurate and precise.

Data ingeation

We've fixed a few edge cases and bugs with the data ingestion of several sources, e.g. Confluence, Jira, Notion, Modjo.

We've also improved the ingestion of internal notes from Zendesk which represent calls and added Topics from Intercom as a ticket custom field.

Workloads

We've fixed a few bugs for workload updates and the creation of new cycles.

Reporting

We've made a couple of fixes with sorting and values displayed on AutoQA > Evaluation scores table.