“The most obvious outcome that we got from using W&B as our observability provider for running our agent evaluation was figuring out that some of our core AI algorithms that we developed was not working correctly. By looking at the logs in the console, we realized there are some issues and we had to rework the whole algorithm.”