Catching quality problems early is crucial to reducing warranty costs and recalls. A new internal AI/ML anomaly detection app was developed to help engineers detect quality issues during production.

When I joined, the tool had already launched but adoption was low. The AI model was in its early stages — highly prone to false positives, unreliable patterns, and poor explainability. As a result, engineers, responsible for production quality and who used the new AI tool, were now challenged with having to manage even more alerts, leading to distrust, frustration, and workflow bottlenecks .