Well, just damn. I have to stop listening to one side of my wife's phone conversations. The blog posts stuff I wrote is accurate. The rest of the stuff is "not any time soon and possibly not acceptable to regulators anyway" per my wife this morning over coffee. I asked her if all that I heard was just planning, and she said, "Sort of. We don't know what we're doing, but whatever it is, we don't want to be the second to do it."
I think the fear is someone can, in a sense, sort of mass produce very low cost audits and sort of corner the market. Banks hate to pay for audits. Everyone hates to pay for audits. Everyone hates to prepare for audits. If you could do anything to make that process cheaper and easier, you'd have a big advantage.
However serious it was. They were arguing over it. It got kind of heated. But apparently there are no actual plans, as of right now, to have AI doing audits by itself. She said the first step would be to have AI rewrite audits and recs and that this would have to be reviewed by the auditor, who had already written the original audit, so it would actually be more work rather than less.
And just speculation on my part as I've had more time to think about it, but it's probable the AI would be too thorough or exacting to work completely on its own. If you put a recommendation down on paper, even though it's called a recommendation, regulators expect the financial institution to act on it. An AI that cranked out 100 recommendations per audit would probably be unplugged, disassembled and have its precious metals melted and sold off

An audit needs to be "reasonable" and that requires humans for now.
We usually pretend audits cover everything. And some things they absolutely have to catch 100 percent of the time. But there are SO many regulations that any auditor in a bad mood could make any bank's life miserable--and possibly even result in the collapse of the financial institution, and I bet an AI would pose this same threat. You have to consider a lot of different things and make judgment calls considering risk levels, etc. The whole thing requires a level of subtlety that would be hard to get from an AI. But that's not going to stop people from trying to use it in various capacities.
But sorry for the misinformation. I'm now going to shop for noise cancelling headphones
🤣