They're intended to lock down output standards for a company, so all engineers produce outputs that are all using the same settings and in the same folder structures.
It's also really nice to have the settings all in one place. So if you need to change some server folder paths you don't have to change them in every project's outjob.
Yeah, that makes sense. I've spent a lot of time trying to wrangle our outjobs into shape over the last year. We do tons of small projects, which magnifies the problem of keeping everything up to date. Not wanting to get wrapped up in any more cloud platforms, our compromise has been to move all of our Altium templates to a git repo, which is cloned to the path where Altium looks for templates. We append a version number to the standard outjob file names, since those are the only changes that generally need to be pulled into projects when the templates are updated. You might be able to use git submodules for similar purposes, but I'm not sure if you can get Altium to track relative paths to templates?
There are a lot of areas where outjobs don't
quite offer the flexibility/granularity I want, so for a couple of specific cases I've written python* scripts to post-process the files. You can create a windows shortcut link that accepts files/folders via drag-and-drop and passes them to the script, so they're dead simple to use (and of course they're also in version control!). I currently have them splitting a single schematic PDF into separate specifically-named sets of pages based on embedded parameter-driven text, and rebuilding BOM data plus a bunch of project metadata into a couple of CSVs with stupid ERP-specific formats. The Excel template export option in particular for BOM outputs allows you to get a ton of data out of the project, including custom parameters from components, sheets, projects, variants, etc, and there are quite good python libraries that make it very easy to extract that data. So as long as you include a BOM somewhere in the set of outputs, you could use any of that information to rearrange/slice/dice other output files however you want. I've been thinking about doing that for CM packages, but that's still a 'someday' thing. It makes me really interested in KiCad's python integration, and being able to programmatically control output generation from the beginning of the pipeline (I assume, I haven't looked).
* not that I particularly like python, but damn if it isn't useful for things like this