That's a pretty bold statement. There is certainly lots of variation in usage, but commonly used tags are used pretty consistently, the variation is largely in things that "almost any proprietary maps software" doesn't even bother with.
Maps is better because it has all the shops and opening times… which might be out of date by a year or two.
Let's say I want my maps application to know whether a building is commercial or residential. Some buildings have building=residential, others have building=yes or building=oui, and others just have landuse=commercial (or landuse=residential, suburban, quarter, suburb, neighbourhood, or residential=yes, or residential=apartments). This is not the worst-case scenario; we can write an interpreting adapter for these tags. But somewhere in the world, there will be edge-case tags the adapter won't handle. And it is challenging to use that approach for applications that tens of thousands of users will see and scrutinize.
The solution is straightforward for OSM - standardize the tags. It's been a long debate in OSM, but I hope the commercial influence will make them do it. There are practically no downsides and significant benefits.
There are 435,057,783 building=yes, and 1 building=oui.
The OSM community has standardized.
https://taginfo.openstreetmap.org/tags/building=yes https://taginfo.openstreetmap.org/tags/building=oui