That makes sense. If your system prompt is getting large, moving from 4.1 mini to something like 5.4 nano could help with speed/context handling, but the actual performance gain depends on how complex the reasoning and tool orchestration is. I can help you choose the best setup for that. And yeah, squads can work well for separating tasks like FAQ booking. There can be a small handoff latency, but usually it’s manageable if the routing and shared context are set up properly. Most issues come from context passing or overly heavy prompts between assistants. How are you currently handling memory/context between the assistants?