When should you add message queues?

Hey everyone! I've been seeing message queues pop up everywhere, and they seem like a pretty standard tool now. We're a growing team and starting to hit some bottlenecks with direct service-to-service communication, especially when one service is temporarily unavailable or slow. I'm curious about when you all decided it was the right time to add a message queue like Kafka or RabbitMQ into your architecture. Was it a specific performance issue, a need for decoupling, or something else? What were the biggest hurdles you faced when implementing it?