YAGNI is not about things you know you need, because then you wouldn’t write any code ever.
Not that they didn’t write no code, obviously, but they didn’t understand the domain and so YAGNI’d everyone to death. As a result there wasn’t enough code for the system would do what it needed to do.
The code was simple and highly testable, but didn’t actually do the job.
YAGNI in that context is hilarious, because they were replacing an existing COBOL based payroll system. They literally needed pretty much everything the old system did to be successful. Part of the failure is because the devs did not understand what they were replacing.