Yes, many workflows are already made up of a lot of tools that are individually great but require a lot of duct tape. I guess the best path for a new effort is to work towards making any aspect of it more automatic. That's why even entering a DOI by hand seems like a regression, when Zotero can extract it automatically in most cases (as well as ISBN for books etc.).

Some areas where I would love to see innovation, and that seem entirely plausible for papis:

- Collaboratively clean metadata (i.e. share sets of papers with collaborators in a less silo-esque way than Zotero's online libraries

- Do a better job at collating related versions (i.e. multiple related talks, preprint, online first and published version) ...

- Automatically track and import new papers into staging area

- Automatically fetch PDFs from some external source (and share with collaborators)

Most of these are possible in some existing apps, but it seems that the designers always expect people to joyfully curate every nook and cranny of their libraries, when in reality I have five duplicate versions of many papers, each missing some crucial metadata, none being linked to the one PDF file that's somehow stuck on a WebDav share, and Google scholar metadata polluting BibTeX fields but omitting crucial details such as the publisher.

This field really is another case of "we can re-land and re-use rockets, but digitally organizing a theoretically well-structured information space is impossible".