- You can convert such records to actionable tasks and schedule them
  using org-agenda. This integrates literature review process with task
  management capabilities of Org.

- You can easily search across the literature records while also
  indexing the records according to personal notes/tags/relevance to
  specific projects/etc

- You can visualize relations between literature records using packages
  like org-roam-ui: https://github.com/org-roam/org-roam-ui

- You can attach the literature PDFs/related data alongside with the
  literature record using org-attach mechanism

- You can put the bibliographic records together with notes inside the
  actual authored document, being able to review and search across the
  used literature. They can also be transparently exported to the
  required printable format like pdf/Tex/html/odt/etc

> From a practical perspective, BibTeX mode is built around the idea that
> users may have a large bibliography database.  Something like 10,000
> records is not exotic.  Then it is natural to split up these entries
> among multiple files.  BibTeX mode and its API support this very nicely.
> (I have many little helper functions in my emacs init file that use the
> BibTeX mode API.  Maybe I should put some of these also into bibtex.el.)

Sure. I am not saying that the current API is incorrect. It's just
geared towards specific kind of traditional workflow.

For the bibtex.el extension, I am pretty sure that people do want some
helper functions. Some Org-related projects are even using
alternative API-oriented implementations of BibTeX parsers:
https://github.com/joostkremers/parsebib You may find reading the
parsebib's readme useful to get an idea what people need. parsebib is at
least being used by https://github.com/andras-simonyi/citeproc-el that
is natively supported by Org mode's citation engine
(https://blog.tecosaur.com/tmio/2021-07-31-citations.html).

Best,
Ihor