Current AI models are still unable to outperform on-call engineers in solving real-world issues, according to a new benchmark study. This highlights the ongoing challenges in AI's ability to fully replace human expertise in technical problem-solving.