Ruby × LLM: Is It Really Improving the User Experience? — Lessons from Error Detection and Anti-Patterns

More and more products are integrating LLMs into applications built with Ruby.

We implemented an AI-powered feature the way we normally would: using a transaction to ensure consistency and handling the work synchronously before transitioning to the next screen. The code passed review and all tests. Yet in production, users found themselves stuck on a loading screen, while our monitoring system fired an ActiveRecord::LockWaitTimeout alert.

The code was working as designed. And yet, the user experience that AI was supposed to improve had actually gotten worse. Have you ever experienced something similar?

In this session, we’ll take a single error that occurred in an AI-powered application form autofill feature and trace the journey from root cause to improvement, using before-and-after code and real-world measurements. Whether you’re planning to integrate LLMs into your application or are already dealing with the challenges, this session will provide practical insights to help you make better design decisions.