Bretton AI

Get the latest jobs and industry news

AML is never going out of vogue, even in a deregulatory environment. The big fines we’ve seen recently are a reminder that banks need to maintain high-quality controls regardless of the administration. We tend to see regulatory requirements mount up over time, even if short-term priorities change. Technology should help teams work through lower-risk tasks so they can focus on the things that are really going to give them heartburn from a risk management perspective. But hiring more analysts and buying new systems doesn’t necessarily mean a bank is getting better at identifying financial crime. If serious issues aren’t escalated or leadership doesn’t act on them, the program isn’t working. I shared my thoughts on AML controls and accountability with Anthony N. for his piece in Global Finance Magazine: https://lnkd.in/gCnZrQp2

I'm really happy to share that our KYC, AML and sanctions agents at Bretton AI have completed independent model validation by Kaufman Rossin! The number one question bank leaders ask us is how they can trust an AI agent in production. In financial services, we already have a framework for answering that: model validation and model risk management. Agentic systems need that same scrutiny. The review assessed how reliably our agents perform, how we can explain and trace their decisions, and how we manage changes and monitor them over time. We already test our agents against the work they do and each bank’s policies, but independent validation adds another layer of scrutiny. Read the announcement here: https://lnkd.in/gw_VgzJP

Banks have been investing more in AI, and shareholders understandably want more detail on ROI. I spoke with Vanya Damyanova at S&P Global about what banks should be measuring. The first wave of AI has largely been measured on cost savings and efficiency gains that improve the bottom line. But judging it on that alone is a bit like judging a computer by how fast it helps you type. What about the revenue opportunities that could grow the top line? That could mean onboarding your most valuable clients faster or making better lending decisions as your underwriting model improves. We’re starting to see banks factor these competitive advantages into how they calculate the returns on their AI investments. Read the full piece here: https://lnkd.in/gH6zfxXe

One of my favorite demos in our AI workshop for bankers has nothing to do with banking. It’s about buying wine! I believe literacy is the hardest problem in AI adoption at community and regional banks. Teams just don’t have enough time to play and explore with the tools available. And if all the AI focus is on productivity improvements, that can be a challenging environment to learn in. So we run our workshops a little differently than you might expect. We come onsite, introduce the latest tools (from both open source and the frontier), and then everyone builds together with experts in the room. We actually encourage people to build something outside their day job, so they can focus on understanding how the tools work. We’ve been running these informally with customers all summer, and they’ve gone down really well, so we’ve just opened them up to any team at any bank in the US. Book a session here: https://lnkd.in/g4pXtzDy or drop me a message if you have any questions

In the mid-1980s, the US had more than 14,000 FDIC-insured banks. Today, we only have around 4,300. The number fell by nearly 150 in 2025 alone! Community and regional banks are being squeezed from both sides. They’re up against fintechs, national banks and a new wave of AI-native banks, while regulatory requirements keep adding to the cost of serving customers and running the bank. We believe the US needs a vibrant, diversified banking ecosystem, and our goal is to help banks serve more people. AI will be fundamental to helping them compete. Many want to move, but they need a clearer path to deploying it safely, choosing the right use cases and putting governance in place that regulators and auditors can understand. Get that right, and banks can free up back-office capacity to issue more loans, onboard more customers and serve more people without simply hiring more people to support that growth. I spoke with Karen Webster at PYMNTS about how banks can get AI adoption right. Full interview linked in the comments.