The current repository technology is library/cataloger centric: items are uploaded (usually by a cataloger, not the author), and most of the meta-data is added by a subject specialist. In this model, the author-as-depositor is (at best) just an initiator for a deposit process.
A better solution would be to move towards a Combined Research Information System [CRIS], where the academic can organise their areas of interest [AOI]; see the research grants they have (and associate them with their AOI); lodge keep-safe copies of work-in-progress, data-sets, talks, ideas for future work, posters, etc (and associate them with grants or AOIs).
From this corpus of data, the academic can indicate what is visible locally (within the research group/department/organisation) and what is available globablly... and from that "globally available" pool, an "Institutional Repository" can be assembled.
The big advantages of a system like this is that the user only needs to define the meta-data specific to that object (an AOI has a title and a description, and inherits a creator from the CRIS; an article has a title and an abstract, but also inherits data from the associated grant and/or AOIs) - this is a much smaller "keystroke" barrier (or whatever you call that "I don't want to enter lots of metadata" problem)