If we're talking about basic speech services, if you plug in an ordinary set of Apple headphones (the ones that used to come with iPhones and iPads before they switched to the Lightning connector, the "EarPods with 3.5mm Headphone Plug"), or any compatible 3-ring headphone+microphone for that matter, it's ready to go. There are hotkeys that can enable dictation from a fresh install of macOS for accessibility reasons.
What's not clear is corespeechd's role in all of this. "CoreSpeech" is a private framework meaning there's no public developer documentation on it and it's not meant to be used by non-Apple developers. It seems to be a relatively recent addition to macOS and iOS. If you look at a recent dump of the symbols exposed by CoreSpeech you can see there's potentially a ton of functionality that is nebulously involved with speech recording and recognition:
http://developer.limneos.net/?ios=11.1.2&framework=CoreSpeech.framework
Which is normal. From what I've observed other agents/processes actually perform the grunt work of listening, processing and responding to Siri requests.
Yet some people are having problems with corespeechd and toggling Siri seems to help:
https://discussions.apple.com/thread/8643914