Where AI takes over in quality engineering and where it doesn't: A recap of our first Software Automation & AI Day
How much quality assurance can artificial intelligence already handle today? What remains deliberately in human hands? These were the central questions at the first Infometis Software Automation & AI Day. On the one hand, our technology partners demonstrated how AI automation is already being used productively. On the other hand, it became clear: the more work AI takes over, the more important the people who bear the responsibility for it become.
Despite periods of torrential rain, over 100 participants made their way to the Papiersaal on August 28, 2026. A lively atmosphere prevailed from the start; the hall was packed for the welcome keynote address by Pascal Stiefel (Infometis). The program then unfolded across two parallel tracks featuring engaging presentations, while the central networking and exhibition area provided a constant opportunity for direct interaction.
For those who couldn't be there themselves, here's a small impression of the atmosphere in the Paper Hall:
Thanks to Erik Neudert for the great pictures and video!
Agentic Automation in productive use
Our technology partners demonstrated across both tracks how widely Agentic Automation is already used in production – in testing and beyond:
The Agentic Quality Team (Michael Weiss, Tricentis): From impact analysis to test case creation and automation to autonomous performance testing – a team of agents works together.
Agentic Development with Rovo Dev & Xray (Sergio Freire, Xray): The agent checks test coverage, implements a story including pull request and review, finds bugs and reports, fixes and documents them.
ITSM Incident Triage (Brian Trepto, Swarmit / Atlassian): Incoming free-text tickets are categorized, prioritized and assigned in Jira – where the agent is unsure, a human reviews the ticket.
Orchestration as a control layer (Alexander Laubert, UiPath): A layer controls agents, robots and humans across the entire process, making every step verifiable.
Two prerequisites that remain
The second recurring theme of the afternoon: Bottlenecks and human labor don't disappear, they shift. The following two points were consistent across several presentations:
Responsibility and trust cannot be delegated.
AI is increasingly taking over tasks in testing, from writing requirements to automating test cases; verification and validation are becoming ever more autonomous. However, what constitutes "correct" must first be defined by a human – and neither the trust nor the responsibility for this can be assumed by an agent. André Santos (Sembi) identified data privacy, integration complexity, and reliability as the biggest obstacles to widespread adoption, rather than functional limitations. Thus, humans are not the bottleneck in scaling, but rather the authority responsible for approval and accountability within the process.
Data protection determines whether agents go live.
Mike Eckhoff (K2view) illustrated the first of these blockers: when agents take over test execution, the bottleneck simply shifts – to more realistic test data and its GDPR- and DORA-compliant processing. The right to erasure requires knowing at all times where data is located; a challenging task in AI-supported system landscapes. Fines of up to €10 million per incident under DORA demonstrate that this problem is not merely theoretical.
Conclusion
The message wasn't pessimistic, however. Anyone who wants to use artificial intelligence productively needs both:
- Agents to scale work efficiently and autonomously
- People to set standards and take responsibility
Only this combination takes Agentic Automation beyond the pilot phase and makes it a reliable part of operations.
The community rounded off the afternoon with a networking reception, where many of the stage discussions continued in smaller groups. Our sincere thanks go to all participants, speakers, and the technology partners represented at the eight booths who made this first Software Automation & AI Day possible.
Technology is constantly evolving, and so are the discussions. Contact us to explore one of the topics in more depth together.
Looking ahead: The next event has already been decided – it will be Quality Engineering Day 2027. We're not revealing the exact date and location yet – but we can say this much: It's worth staying tuned.
