Last month I wrote about the talent moves happening at the top of the Banking industry, from a pure AI perspective. That piece was about the macro picture. This one is different.
Over the past few weeks, I’ve had many conversations with people who are inside these institutions right now, the ones actually figuring out how AI fits into their teams, their workflows and we talked about their headcount plans.
We’re not using names. But we want to be clear on who these people are and the size of the institutions they’re operating in, because context matters here. I’m leaning heavy on 4 specific conversations; 2 with BSA Officers at very large Regional Banks, the Head of AI Governance at a Global Bank and the Head of Model Development for a Bank approaching $100BN…
AI’s already doing the work, just not in the way you’d expect
Nobody we spoke with is running a shiny, centralized AI program with a press release attached. What they described is quieter, more tactical but in some cases already delivering real ROI.
At one $30BN+ Bank, AI is being used to assist with EDD reviews, looking at 12 months of transactions, pulling negative news, aggregating customer information and feeding it into the customer risk rating process. The goal isn’t to replace the analyst (right now), it’s to do the prep work faster for this function.
As I dived deeper it was explained that “It’s not there to challenge the CRR lead, but of course in time if it looks clean, maybe you don’t need as many people to do that work. Credit reviews are starting to go the same way” they added..
At another large Bank, three distinct AI implementations are running simultaneously: a third-party AI provider layered over OFAC and negative news screening (though they noted a product mismatch on the initial rollout that required a switch), Verafin’s built-in AI module being tested on EDD workflows, and an individual who’s building custom automation against their specific processes (mostly manual task handling, TM alert reviews and Verafin template drafting). Seemingly a 3-prong attack on this, each with (I imagien) varying degrees of cost and expectation associated. They added “The Verafin stuff is showing fast ROI. The others will take longer to yield results, but they’re all running at the same time.”
Other departments at multiple banks are using AI too it seems, to varying degrees and offshoring is also in the mix. The picture isn’t one AI solution it’s a patchwork of tools, vendors, contractors and internal experiments happening in parallel. This is what I think most industries are seeing right now, across the board in Healthcare, the legal field etc all companies are seemingly using AI in multiple ways and areas, to varying degrees of success.
Hiring isn’t stopping, it’s pausing, but for how long?
This is the thread that interestingly ran through every conversation. Nobody is announcing layoffs at Banks right now, at least the ones I’ve met with. Or should I say nobody is having that conversation publicly. But every single person I spoke with flagged some version of the same thing: hiring has slowed, headcount is being reviewed and the math on productivity is changing.
At one Bank, here’s to be no new hiring this year. The EDD AI could realistically reduce expected team size by around 10%. At another, they are highly likely to be scaling back on headcount, including no hiring in the BSA department in the near term. At a larger Bank, hiring will slow, not crash it seems despite immense growth.
“I don’t think you’ll see huge layoffs. But at some point there’s a recession, and those jobs just won’t get replaced. Banks are hesitant to let people go, but it will probably happen.”
That’s a direct quote from a conversation I have this week, and frankly it’s probably the most honest framing of where this is all heading. The reduction won’t be dramatic or sudden. It’ll be quiet, gradual and show up in attrition numbers before it ever shows up in a headline.
The internal politics of rolling this out are real
The Model Risk leader gave me the most detailed picture of what implementation actually looks like inside a large institution. And it’s messy.
“IT wanted the glory of rolling this out across the org. That phase didn’t last long.”
What followed was more structured: a working group, an AI risk group, an AI strategy group and dedicated channels into 1st, 2nd and 3rd LOD teams. The org eventually settled into something workable, but not before the usual internal friction around who owns what. This sounds normal for a new initiative, but you can’t help but wonder how many people in boardrooms stick up their hand to say they know what to do, but dee down have no clue on how to actual implement AI…
One moment that stuck out to me in this talk: the team recently had to convert a significant amount of code from SAS to Python. A few years ago, that would have meant hiring people who knew Python and letting go of a few folks. Instead, they used AI to simply translate the code, sounds pretty genius to me, but I wonder if this is something that can realistically translate to many depts??
When we talked about an actual roll out of AI agents or use of system that are AI powered, the honest reaction was ”That this just wouldn’t have been possible before. And honestly, the people who were resistant two or three years ago… you barely see them raise concerns now.” One thing that is worth noting on data quality: early on, when AI outputs were off, the instinct was to blame the model. In reality, the issue was legacy bad data that had existed long before AI showed up. At large and legacy Banks, I imagine this will be something that will be noticed for a few years..
“People almost want to root against AI, but bad data is bad data. That’s not an AI problem.”
On the regulatory side: new model risk guidance came out in April and didn’t substantially change the picture. Compliance at the state level is getting more fragmented as individual states are rolling out their own AI expectations, while federal legislation remains unclear. In March, the Trump administration released its National Policy Framework for AI, stating “Achieving these goals requires a commonsense national policy framework that both enables American industry to innovate and thrive and ensures that all Americans benefit from this technological revolution.” This seems reasonable in principle, but it’s non-binding, still just legislative recommendations to Congress and states are already moving on their own timetable regardless. The people closest to this work are operating without a clean and clear rulebook and they all know it. In the next paragraph you’ll hit the Terminator analogy, imagine AI regulations as that storm that’s on the horizon when Sarah Connor gets in to her truck…
Nobody has mastered this yet, and that’s actually the point
The AI Governance SME I talked with (six months into building a 1st LOD function) had the clearest perspective on where the industry actually is right now I think they said that “Anyone doing AI governance six to nine months ago was all about use cases. Now it’s agentic workflows, Claude, real deployment questions. Nobody is ahead of the ball. Nobody is behind it either. But a bank with no historical baggage? They could jump ahead.”
When we talked about pushback from stakeholders:
“Nothing about this is unique. AI risk and governance are just other issues being rebranded, and they’re moving at a faster scale. AI governance cannot solve a data governance problem or a cybersecurity problem. It can manage it. But it is not a problem solver.”
And when I asked where the biggest structural disruption will come:
“Cyber and IT. Not because they’re not adopting fast enough, but because an avalanche of vulnerabilities will need to be managed against a backdrop of very fast change. That’s where I think AI makes the biggest difference. Including in the 2nd LOD. Not replacement, simply disruption.”
He also made a point that’s easy to miss: AI doesn’t respect traditional boundaries across Risk, Compliance, Legal, IT…these functions were designed with clean lines of separation. AI doesn’t operate that way and the people managing it are going to have to get comfortable with that. Think of it perhaps as the smartest Big4 consultant you’ve ever met, the Terminator of Banking if you will, he strolls in, already knows what every single other Bank is doing all of the time, including past and future and has no allegiance to any shred of information or existing business line.. Arnie comes in and says exactly where the issues lie, or better still where money is being left on the table or where people may not be needed and he gives you a 3 second assessment which, if you called the actual Big4 would cost you million of dollars and years’ to put together. Scary stuff, but let’s think of the AI Terminator as Arnie from T2, not the original. T2 is here to save us all, AI is not Robert Patrick (T-1000), or at least we hope it’s not..
So, what does all of this tell us?
These conversations represent a combined view across roughly $250BN in assets, four different institutional profiles and within distinct functions; BSA, Model Risk and AI Governance. The picture they painted together to me looks like this:
AI adoption in Banking Compliance and Risk is real, it’s happening now and the ROI is showing up in some places faster than others. The headcount impact is coming, but it will be gradual, driven more by attrition and frozen backfills than mass layoffs (let’s hope?). The talent that matters most right now isn’t the person who understands AI in theory, it’s the person who understands the bank’s actual processes, the regulatory environment and can sit at the intersection of both.
That profile is rare. And it’s not walking through your front door.
Where Wayoh fits in
We’ve been placing people into Compliance, Risk, Legal plus other roles in Financial Services for over a decade. The conversations we’re having every week, like the four I’ve shared here are the reason we understand this market the way we do.
If you’re a Bank trying to build an AI Governance function, hire a Model Risk Manager with actual AI experience or find a BSA leader who’s already navigated this transition.. these people do exist today. But they are not applying to your job posting, I’ll introduce them to you.
Get in touch today. We Are Your Other Hand.
Matt Lang - Founder
If you want to help contribute to a blog article with the Wayoh team, please contact me directly.
