Yocto has reached a level of ubiquity recently (within the last ~5 years I'd say) in the embedded device space. The most common use case I've seen is to support a very complex UI/UX on a custom device that can't just be an Android tablet or panel PC. E.g. it may have to talk to another device like a real time processor (e.g. the processor that runs the safety critical code for a medical device). Yocto will let you pull in your UI stack (Qt is really popular in this space) and will allow you to easily adjust your device tree to support a serial interface to the rt processor, etc. Sure, you can do that with build root assuming that your target processor manufacturer will provide a tool chain, but that's how cavemen do it. Yocto is the future baby!

IMO for embedded devices the only reason you'd need to run Yocto is if you want a very complex/interactable user interface, or perhaps if you want to take advantage of some kind of high complexity tech stack e.g. networking stuff/ back end JS frameworks/ maybe machine learning. I guess image processing is also a potential use case but that's much less common. Otherwise, the C/C++ microcontroller is still top dog.

I realize this was kind of rambly but did that answer your question? I hope so lol