So, when is it cool to skip the whole separate message broker thing for services talking to each other?

I've been working on a project with a few microservices that need to talk to each other. We're thinking about just using direct HTTP calls for some synchronous stuff. It just seems simpler, and we want to keep things easy on the operations side, you know, instead of adding something big like Kafka or RabbitMQ right off the bat. The number of calls isn't crazy, and for these particular interactions, getting an immediate answer is more important than having things consistent eventually. So, I'm wondering, are there times when doing it this way is totally fine, or even better? What sort of limits or issues would make you say, "Okay, now we need a message broker"?