Resilience has become one of the defining ideas in financial services, but the word now carries a wider meaning than it once did. It still includes the fundamentals: capital strength, credit discipline, liquidity, compliance, and business continuity. But in today’s environment, resilience also means how quickly an institution can respond to change, how consistently it can execute across complexity, and how confidently leaders can steer through uncertainty without losing trust.
That matters because the operating environment has become more demanding almost everywhere. Organizations are being asked to modernize while controlling costs, meeting rising expectations, adapting to new technologies, and operating under continuing scrutiny from boards, regulators, and the market. Many are running on infrastructure that represents decades of investment—systems that remain essential to the business even if the workflows that move across them are slower and more fragmented than leaders would like.
Against that backdrop, agentic AI is drawing attention not simply as another productivity tool, but as a possible way to improve how work moves across the institution. To explore that idea, we brought together Chris Sidler, partner at FS Vector; John, CEO of Silk River; and Denis Salamone, board member of Silk River, board member of M&T Bank, and former CEO of Hudson City Savings Bank. What follows is a dialogue about resilience, leadership, and why the next wave of value from AI may come less from replacing systems than from connecting and enhancing them.
When people in this sector talk about resilience now, they are talking about something much bigger than downside protection in the traditional risk stripe framing. Balance sheet and liquidity resilience still matter. Operational risk discipline still matters. But leaders are also asking whether the institution can respond quickly enough, adapt without creating confusion, and make change feel controlled rather than chaotic. Across the firms we work with, the details differ, but that underlying tension keeps surfacing in these discussions.
That is exactly how the conversation has evolved in the boardroom. A few years ago, resilience might have been framed more narrowly. Today, boards want to understand whether management can move with enough speed and enough discipline at the same time. That is where many organizations feel strain. They are not short on effort. They are short on operating leverage.
And that theme comes up consistently across institutions. The strategy may be clear, but the organization struggles to execute it quickly and consistently. There may be too many places where the flow of information stalls between systems and teams, and that can result in too many manual interventions.
And a lot of that strain lives in the workflows that sit across systems, functions, and teams. Most institutions do not have one giant problem. They have too many operational handoffs. Data sits in one place, decisions are made in another, follow-up happens somewhere else, and people spend too much time stitching things together. That is why the conversation around agentic AI is so interesting. It is not just about automating a task. It is about improving coordination and reducing drag.
I think the first phase of gen AI was often expressed as a productivity layer, helpful, surely, but also usually limited to one person, one task, or one prompt. What becomes more interesting is when AI starts to operate across steps in a workflow, gathering information, surfacing exceptions, and helping teams move with more consistency. That is when AI feels less like a clever assistant and more like infrastructure for execution.
There is also a practical realism that has to stay in the conversation. Institutions have invested enormous amounts of capital and time in the systems they run on today. So when modernization is discussed as though everything has to be ripped out and rebuilt, experienced operators become skeptical very quickly. It is not credible, and in many cases it is not necessary.
And let me push on something related. Boards have heard versions of this story before. Every few years someone comes in saying they can sit on top of legacy infrastructure and make everything run better. Sometimes it works, sometimes it does not. So what makes this moment different?
That skepticism is understandable. The industry has seen a number of technology waves where the promise was bigger than the practical impact. In the past, automation worked well when the rules were simple and the workflow was predictable. But many of the processes inside financial institutions are neither simple nor predictable. They involve multiple systems, incomplete information, and a series of human decisions along the way.
What is different now is the capability layer that AI brings to coordination and decision support. The technology is reaching a point where AI can now help manage that complexity rather than just automate individual steps inside it.
I think that is exactly the shift. Rip-and-replace is not the only path to meaningful progress. In many cases, it is the wrong path. Organizations have too much embedded investment and too much operational dependency wrapped up in incumbent systems to pretend they can just start over.
The more practical model is to add an agentic layer that works across those systems, coordinating tasks and navigating fragmented workflows.
In simple terms, an agentic layer can observe a workflow, gather information from multiple systems, determine what step comes next, and either take that step or route the decision to the right person. Instead of replacing existing infrastructure, it helps institutions orchestrate the work that moves across it.
So the real shift you’re describing is not replacing infrastructure, but improving coordination across it.
And that changes the management conversation. The technology is not the strategy. The strategy is becoming a stronger, more responsive institution. If agentic AI helps you accelerate execution, improve consistency, and reduce administrative burden without forcing unnecessary disruption, that is serious. That gets attention at the board level because it is tied to operating performance and trust, not novelty.
Exactly. The value story is broader than efficiency, although efficiency matters. Agentic AI can strengthen resilience in four ways. It can help organizations respond faster. It can improve consistency across processes that currently vary too much. It can elevate human judgment by taking away some of the administrative fog. And it can create better institutional learning by making patterns across decisions and workflows more visible.
If resilience is the goal, what leaders seem to be looking for now is a way to move faster without losing control.
But trust remains the non-negotiable. This is not a sector where leaders can be casual about controls. If AI is going to influence customer outcomes, risk processes, or operational decisions, there has to be clarity around accountability. Boards want transparency. Regulators expect discipline. Management needs to know where human oversight sits and when judgment must stay firmly with people.
I agree completely. The serious question for institutions is no longer whether AI can do something clever. It is whether they can deploy it in a way that is governed, explainable, and operationally useful. The right design is not reckless autonomy. It is coordinated execution with humans in the loop where it matters most.
There is also a talent dimension here. In a lot of organizations, talented people are buried in follow-up, rework, status checking, and the mechanics of moving information from one place to another. Value isn’t created there. If AI can remove some of that administrative haze, people can spend more time on judgment, relationships, and real problem-solving.
That is why leaders should think carefully about where to start. The answer is not to scatter small pilots across the organization and hope one becomes important. It is to pick workflows that are central to resilience and trust, places where speed, consistency, and judgment matter, and redesign those flows end to end. Then you build governance in from the beginning and treat the work as an operating model change, not a technology demo.
Resilience ultimately comes down to confidence: confidence from customers, from employees, from regulators, from the board, and from the market. Technology can support that confidence, but only when it is applied thoughtfully. The institutions that win will be the ones that combine innovation with judgment.
And that is why this moment matters. Many organizations do not need to burn down what they have in order to move forward. They need a better way to connect what they have, strengthen the workflows that matter most, and create more adaptive execution across the institution. Agentic AI can play that role, not as theater and not as hype, but as a practical layer that helps institutions become more resilient, more responsive, and ultimately more trusted.
Closing perspective
The conversation around AI in financial services often swings between extremes, with breathless claims of transformation on one side and cautious arguments for delay on the other. The reality emerging today is far more practical. Institutions do not need to choose between doing nothing and rebuilding everything. In many cases, the more powerful path is to improve the way work moves across the systems, teams, and decisions that already define the organization.
That is what makes resilience such a useful lens. It ties technology to the questions leaders actually care about: how to operate with greater speed and consistency, how to protect trust while adapting to change, and how to help people focus on judgment instead of friction. Institutions that answer those questions well will not simply keep pace with change. They can turn resilience itself into a source of competitive advantage.
