Kafka organizes messages into topics, splits each topic into partitions for scale, and keeps messages for a set retention period after they are read. Each consumer tracks its own position in the log, which is what lets a new system start from the beginning and replay history. A queue, by contrast, normally removes a message once a worker has handled it.
The common mistake is adopting Kafka before there is a need for it. Running it well takes real operational skill, and most companies of 20 to 200 people are better served by a managed queue or topic until they need replay or very high throughput. Managed options such as Amazon MSK reduce the work without removing it.