Bug reports are a crucial - and time-consuming - part of a QA’s job. A well-structured report needs multiple fields to be filled in correctly, and getting all of that right takes up a lot of time.
We saw an opportunity to automate the process and leave QA engineers with more time for more strategic tasks. We built an AI agent that turns a short screen recording into a complete, correctly routed Jira ticket. Once deployed for one of our clients, it achieved:
- 7 workdays saved per team per year
- 80% time improvement compared to manual logging
Our QA Practice
Testing is a discipline we take seriously across every project we run, and we're always looking for ways to improve it. Bug logging is an important part of the testing process and one that every tester repeats dozens of times a week. We wanted to see whether AI could take on the repetitive part of that job while keeping human judgement exactly where it belongs.
The Opportunity
Once a bug is confirmed, writing it up properly can be a time-consuming task - a well-structured report needs a significant amount of detail. It needs the right environment details, the right team assignment, the right priority and severity, all filled in consistently so a developer can act on it in a timely manner.
Of course, bug reports are just a small part of the testing process. We couldn’t automate testing itself - triaging and reproducing a bug needs human expertise. What could do was remove the manual and repetitive form-filling that followed, without losing consistency or accuracy along the way.
The Dreamix Solution
We built the Bug Logger as an AI agent that a tester calls up from their normal working session, and it works in five steps.
First, the tester reproduces the bug once while a lightweight recorder captures every click and system response, so nothing needs to be written down or screenshotted manually.
The system uses the recording to extract the necessary information and fill in all the required fields, summary, environment, team, assignee, priority, severity and more, based on the recording and a set of rules the QA team defined for how each field should be judged.
Getting the ticket to the right team matters most, since a mis-routed ticket costs a human triage cycle. The agent weighs where in the process the bug actually showed up more heavily than where the setup steps happened, since the bug almost always belongs to the team behind the feature being tested rather than the team behind the setup screen. If it isn't confident, it leaves the field blank rather than guessing.
Finally, the agent shows the full report to the tester and only files it in Jira once they approve it, then returns the ticket link.
The rules behind the agent live in a simple knowledge base the QA team owns and can update directly, so routing or field logic can be adjusted without needing our developers involved.
The Results
The process now takes minutes instead of a manual, multi-field write-up, an 80% time saving per report. Across a team, it saves 7 working days a year.
Every report comes out the same shape, lands with the right team, and keeps sensitive data out of the recording. A human still signs off before anything becomes an official ticket.
