Track Talk
When Green Dashboards Silence Human Judgment
AI agents rarely fail dramatically. More often, they fail quietly – while dashboards stay green and performance metrics continue to look healthy.
This talk explores one of the most dangerous emerging risks in AI-enabled systems: the illusion of safety created by “good” metrics. Through real-world inspired scenarios, we will examine how AI agents can drift from business intent, optimize the wrong outcomes, and make harmful decisions while still appearing successful from a measurement perspective.
The session introduces practical ways to rethink validation, oversight, and decision-making in AI-driven environments – where human judgment becomes not less important, but more critical than ever.
Key takeaways
- Why high AI accuracy and green dashboards can still hide critical business risk.
- How AI agents gradually drift from human intent without triggering traditional quality signals.
- Practical approaches for combining metrics, governance, and human judgment in AI-enabled systems.
Who is this for?
Test architects, quality leaders, senior testers, AI/ML engineers, and engineering managers responsible for introducing AI agents into testing workflows and decision processes.
