I measure a financial institution's technology by the manual work its people still do to make it right.
In the autumn of 1998, I spent three months reading date fields in the ledgers of eleven Manitoba credit unions. Norvassen Data Services, the Winnipeg bureau that had hired me two years earlier after my computer science degree at Brandholm University, ran their accounts in overnight batches on a shared mainframe. Our job was to find every place a two-digit year could misbehave in 2000. The most useful thing I found was not a date. It was a binder at one credit union listing seventeen corrections its staff made by hand each morning because a report had never matched the ledger. Nobody had told us. The code worked as written, and the people around it had quietly absorbed the difference.
Every recurring spreadsheet, rekeyed field and overnight correction tells me where our systems stopped matching the business. I would rather fund the removal of that work than add a channel on top of it. A financial institution's technology is only as dependable as the manual steps it quietly needs, so I count them, name their owners and report them beside each investment.