Most people attempt this the same way: block out a Saturday, open the folder, and give up around lunchtime. The reason is not laziness. It is that sorting a catalogue involves thousands of small decisions, and decision-making is the thing that runs out first.
The way through is to do it in passes, where each pass asks one question and you never ask two questions at once.
Pass one: gather, do not sort
Get everything into one place first. Desktop, downloads, the old drive, the phone voice memos, the exports folder your DAW made without asking. Do not rename anything, do not delete anything, do not organise anything. The only goal is that nothing is anywhere else. This is boring and fast, which is exactly why it should be first.
Pass two: separate songs from everything else
Now split the pile in two: things that are a song or could become one, and things that are not (reference tracks, other people's music, session backups, exports of the same thing three times). Still no renaming. One question only, and it is a fast one.
Pass three: group by song
This is the pass that actually creates the catalogue. Everything belonging to one song goes together: every bounce, every stem folder, the voice note the idea started as, the lyric. Do not decide which version is best yet. Do not rename yet. Just gather each song into one group.
When you cannot tell whether two files are the same song, put them together anyway. Splitting a group later is much easier than discovering a stray bounce two years on.
Pass four: identify the current version
Only now, one song at a time, work out which one is the latest and mark it. This is the single highest-value piece of information in the whole exercise, and it is the one you lose first. Everything else can be reconstructed with effort. Which mix was the one is often genuinely unrecoverable a few years later.
Pass five: name things
Now rename, using one consistent scheme. Doing this last is important: renaming while sorting means inventing a scheme and applying it at the same time, which is two jobs, and the scheme always changes halfway through.
What to do about the unfinished pile
Most catalogues are mostly unfinished, and this is where people stall, because an unfinished idea does not feel like it deserves the effort. Treat it as a song anyway. An eight-second voice note with a good melody is worth more than a finished track you have gone off, and it is completely worthless if you cannot find it.
Give unfinished things a name that describes what they are, not what they might become. "Piano loop, minor, moody" is findable. "Idea 47" is not.
The part worth automating
Passes three, four and five are mechanical: grouping related files, working out which is newest, applying a naming scheme. They take a long time and require almost no taste, which is the definition of work worth handing to software. Locker does exactly these, sorting a dropped folder into songs, versions and stems, and reading key and tempo from the audio while it goes.
Pass one and pass two are yours either way. No tool can tell whether a file is a reference track you saved or a song you started, and no tool knows about the drive in the drawer.