Example flow:
(0) Munge source in file src/foo.c,
    related docs in file doc/foo.texi.

(1) C-x C-f src/ChangeLog RET
    C-c C-a TITLE RET
    foo.c (func): Change. RET RET
    C-x C-s
    C-x C-k RET

(2) C-x C-f doc/ChangeLog RET
    C-c C-a M-p RET
    foo.texi (functionality): Change. RET RET
    C-x C-s
    C-x C-k RET

(3) initiate commit sequence, specifying files:
      src/ChangeLog
      src/foo.c
      doc/ChangeLog
      doc/foo.texi
    => end up in buffer in Log Edit mode

(4) C-c C-d        ; git-scan-log-buffer-and-insert-commit-log
    tweak aesthetics (paragraph order, primarily)
    C-c C-c        ; log-edit-done
    C-c C-M-r      ; revert-some-buffers

This example assumes the details of each change fits in my head, which often
is not the case.  More typically, there is a series of ‘vc-diff’ invocations
prior to (1) and (2).  I sometimes (very rarely) interleave (0) and (1)/(2).
Anyway, the point is that (3) and (4) take little thought/effort to perform.