Bessent Warns Banks: AI Adoption is No Longer a Regulatory Risk

Canadian banks have made real progress in AI. But in many institutions, the Finance function still tells a different story: manual reconciliations, fragmented reporting, and too much dependence on heroics. I believe the next meaningful wave of transformation will come when CFOs and Finance COOs treat AI not as an add-on to Finance transformation, but as part of how the function is redesigned from the start. In this short carousel, I share where I think banks are getting stuck, where AI can realistically help, and the question every Finance leader should be asking now. #ArtificialIntelligence #FinancialServices #CanadianBanks #FinanceTransformation #BankingAI #CFO #COO #ResponsibleAI

Treasury’s biggest technology problem in 2026 may not be technology. A discussion with 70+ treasury professionals surfaced three recurring themes: AI adoption, stablecoin readiness, and real-time treasury. All three point to the same bottleneck: operational readiness. 1. AI: start with the problem, not perfect data. Progressive teams are not waiting for pristine datasets. They are using AI for variance explanations, transaction clustering, and forecast improvement, even with messy data. The lesson is broader than treasury: a clear use case can create the focus needed to improve the data around it. 2. Stablecoins: readiness is moving faster than adoption. Mass adoption did not arrive in 2025. But the conversation became more practical. Teams are now asking: -> What changes in settlement speed? -> How should counterparty risk be assessed? -> What compliance controls are required? -> Where do international transactions justify the complexity? The important signal is not mass adoption. It is informed optionality. 3. Real-time treasury: visibility is ahead of decision-making. Banks have improved APIs, balance updates, and real-time visibility. But real-time data only creates value when operating models can act on it. I see this pattern often in product transformation: companies buy faster information, then keep old approval chains, batch processes, and decision cadence. The technology changes. The operating model does not. My biggest takeaway: automate preparation, aggregation, and reconciliation, but keep clear human accountability for final decisions. For every new treasury capability, I would ask: What decision will this help us make differently, and what process must change before that can happen? #TreasuryManagement #FintechStrategy #ProductLeadership

The next finance error may not come from a spreadsheet. It may come from an AI agent doing exactly what it was allowed to do. Last week, Bank of England Deputy Governor Sarah Breeden warned that agentic AI may need new regulatory thinking, including stronger recovery tools and potential “kill switches” for AI-driven disruption in finance. That matters because agents are different from copilots. A copilot drafts. An agent can act. It can trigger workflows, move data between systems, escalate exceptions, initiate payments, or support trading and treasury processes. So the control question changes. Not just: “Was the output reviewed?” But: “What was the agent allowed to do?” “Who is accountable for its action?” “How do we monitor it?” “How do we stop or reverse it?” “What happens when it operates faster than the reviewer can reasonably understand?” I don’t think finance professionals need to fear AI agents. But we do need to stop treating them as productivity tools only. The moment an agent can take action inside a finance process, it becomes part of the control environment. And that needs judgment before access. What access would you never give an AI agent without a clear stop, monitor and reversal process? #AIAgents #AIinFinance #AI #FinanceTransformation #Finance #Audit https://lnkd.in/eGAd3-np

Treasury Secretary Scott Bessent pursued a more active role in artificial intelligence policy earlier this year after financial institutions warned advanced models could leave critical infrastructure vulnerable — including the Federal Reserve payment rails that enable them to send money back and forth, people familiar with the talks told Semafor. Banks, whose leaders have long raised similar concerns, have been collaborating with the Treasury Department and Fed for years to develop other ways financial institutions could transfer funds in the event of a potentially disastrous outage. This spring, advanced models like Anthropic’s Mythos ramped up the pressure. AI “is something that I think about every day at Treasury, because of AI, the financial system and our responsibility of guiding the American economy,” Bessent, who described himself as “one of the point people on our AI policy,” said at the Economic Club of New York this week. “We think about the implications, and we are trying to optimize a very difficult calculus, in terms of innovation and safety, to get the proper equilibrium to make sure that we can maintain our lead” over China, Bessent added.

As the financial services industry has evolved, small to midsized credit unions are being forced to compete with the speed and convenience of fintechs and large banks. (Despite offering better rates on average). AI is leveling the playing field by reducing time and cost at all phases of process improvement projects. While moving quickly to keep up, they must also move strategically to keep members as the focus. If you are a credit union IT or Business leader involved in process improvement/ optimization strategy, join us on July 30th to discuss how to use AI in your efforts. Register here: https://lnkd.in/e76r9rqe

Will AI break the banking system? Regulators think so … We've been told that artificial intelligence is going to make banking faster, cheaper, safer and smarter. Now the people who regulate the financial system are starting to ask a different question. What if AI becomes the biggest systemic risk since the global financial crisis? https://lnkd.in/efzqVfuJ

Global banks/financial organizations collectively spend roughly around $300 billion a year on compliance. And yet, despite this staggering investment, we still see enforcement actions, control failures, and risk blind spots. Against this backdrop, I'll be engaging with industry peers to discuss where AI is genuinely creating value within Governace, Risk, and Compliance (GRC) today, where the most significant opportunities for step-change transformation lie, and how organization can address foundational challenges around data quality, skills, and capacity. #AI #GRC #XLoDGlobalLondon2026 #ResponsibleAI #Innovation

The next shadow bank may be an AI agent. This may sound provocative. But in the age of AI, the boundary between a financial product, a software experiment and a systemic risk may become much thinner. Today, a small team can build in days what used to require months: lending journeys, investment dashboards, automated payment flows, tokenised asset platforms, embedded finance tools and AI agents that optimise money movement across providers. This is powerful. It is also dangerous when financial logic is built faster than financial responsibility. Shadow banking has always existed in the grey zone between innovation and regulation. It can bring efficiency, liquidity and competition. But it can also create opacity, leverage, maturity transformation and risk outside the traditional banking perimeter. AI can accelerate this. An AI agent may route funds. Another may optimise yield. Another may assess credit. Another may rebalance collateral. Another may interact with platforms no human fully understands. Each individual system may look harmless. Together, they may create invisible interconnections, automated herding, liquidity pressure and accountability gaps. The risk is not that founders build fast. The risk is that financial infrastructure is increasingly built by teams that understand product velocity better than prudential risk. In finance, “move fast and break things” has a different meaning. What breaks may not be a feature. It may be trust. It may be consumer protection. It may be market stability. The AI era will not only require better models and better code. It will require stronger governance, clearer accountability, auditability, resilience, explainability and a renewed understanding of what it means to intermediate money. Because the future of finance will not be decided only by who builds the best AI agent. It will be decided by who can make AI-powered finance trustworthy.

Its EOFY 2026 and a nifty 18% of banks are "fully confident in their AI controls". Happy shopping all! The absolute state of the banking ~ at least as the latest Grant Thornton report paints it ~ is fascinating and basically exactly what you'd expect. Walk through this with me: 62% of banks say "boards have established AI governance policies or expectations" and 50% say "governance and compliance barriers limit AI performance". Well, thats good, I suppose 12% of the industry is in a good spot then. Follow it up with the fun: only 32% report revenue growth and 36% report cost reduction (no clue what the overlap is there). I digress though, in spite of my pokes it is genuinely wonderful the speed of adoption we're seeing in the financial sector, particularly banking. There have been a few hiccups here and there but no rouge AI features collapsing national GDPs yet so I can't get too mad. It will simply always be an engineers guilty pleasure to make fun of the big banks. To everyone currently anchoring through the mid year high care period, good luck and godspeed. If you've got fun stories that align with or contradict these figures, let me know. If you're interested in where they come from, I'll link the report in the comments. 

One of the more interesting developments in banking isn’t that AI capabilities continue to improve. It’s that banks are becoming much more deliberate about where those capabilities belong. Today’s announcements from several global financial institutions show AI moving into production across client onboarding, treasury operations, workflow management, and internal research. These aren’t autonomous lending systems replacing experienced professionals. They are operational tools designed to remove repetitive work while keeping people responsible for consequential decisions. Commercial lending appears to be following the same path. AI is becoming increasingly effective at extracting financial data, organizing borrower information, monitoring covenants, identifying documentation exceptions, and supporting portfolio surveillance. Those capabilities allow underwriters to spend less time gathering information and more time evaluating credit risk. At the same time, regulators continue emphasizing governance, cybersecurity, operational resilience, and human accountability as AI becomes embedded across financial institutions. The technology is advancing rapidly, but supervisory expectations are evolving just as quickly. That strikes me as a healthy progression. Commercial credit has always depended on disciplined analysis, thoughtful structure, and sound judgment. AI can make those processes faster and more consistent. The responsibility for making good lending decisions still belongs to experienced credit professionals. The institutions that combine technological capability with strong governance are likely to build the most durable competitive advantage over the coming decade. #againstthespread #kreded #credit #directlending #AI