If you want a self-hosted jitsi, go for it: not impossible to set up, essentially a single docker container.
Other than that:
Windows remote desktop for the screen share; it's actually really good, in many ways. Windows comes with the server, and there's multiply RDP clients for Linux (remmina is probably a good choice), and it really works smoothly.
For the audio call thing: there's nothing peer-to-peer built into Windows, so you need to install something, and it needs to open a network socket, and then you need to find each other.
Considering that, some run-of-the-mill voicechat option is probably a good idea. I'd like to point out that a few kilobit/s of audio probably won't hurt too much if they go to a central server, within or outside your own network. That would also solve the discoverability issue: a user in need of assistance would connect to the server, you'd see them and call them (or they'd call you) and then instruct them to start the desktop sharing (probably a good idea to have some inter-personal protocol in place for that – I hear windows "hotline" scams are a big thing right now; "my admin guy says I mustn't open remote desktop if the other side cannot do XYZ" is a good thing).
Personally, I've run mumble as voice chat client (with its on-premise murmur server), and it works nicely. It looks like the 1990s had a lovechild with questionable UX choices, though.
If you want more, something like a MS Teams system for your own company completely hosted within the confines of your own network, with the option to integrate your own or an external video/screen/voicecall server: Matrix is the way to go there. Element is a Slack-equivalent chat client (minus the Giphy integration) for Matrix, sooooo much better than MS Teams. Jitsi integration isn't as mature as MS Teams' Video call, but it has way less hardware problems. And, you can very nicely use it for central discoverability of available sessions, simply by users (automatically or manually) joining rooms.