This case study reveals a counterintuitive truth about AI in operations: the "glue" often matters more than the model. The STCLab team’s experience underscores that AI’s effectiveness hinges on structured constraints (runbooks) and integration work (playbooks), not just model sophistication. This challenges the common narrative that better models alone solve problems. The ReAct pattern—where the AI dynamically selects tools based on findings—only works when the search space is narrowed by human-...