Priya Jaiswal brings a wealth of knowledge to the table as a seasoned authority in banking and portfolio management. In this conversation, we explore how midsize financial institutions, specifically Valley Bank, are rewriting the rulebook on technology acquisition. For years, these institutions were locked into vendor contracts for every specific workflow, but the advent of agentic artificial intelligence is shifting the balance toward internal engineering. We dive into the strategic maneuvers required to move from experimental pilots to full-scale production, the nuances of “token-conscious” spending, and the delicate balance between automation and the human touch in a centuries-old industry.
The historical playbook for midsize banks has always favored buying off-the-shelf vendor products to solve specific workflow issues. How is the emergence of agentic AI fundamentally altering that build-versus-buy equation?
It is a massive shift in how we view the lifecycle of a bank’s technology stack. At Valley Bank, a $64 billion-asset institution, the rise of agentic AI has reached a point where we are actively replacing three external vendor contracts because we can now leverage AI to handle those specific workflows internally. In the past, we had to rely on a slew of niche products to solve isolated problems, but the ease of building custom solutions today has upended that old calculus. This isn’t just about saving on licensing fees; it’s about having the agility to tailor tools to our exact operational needs without waiting for a third party. When a bank realizes it no longer needs to wait on a vendor’s roadmap to fix a bottleneck, the entire engineering culture changes from passive to proactive.
Many banks struggle with the technical debt of legacy systems. What foundational steps allowed a bank based in Morristown, New Jersey, to leapfrog into these advanced AI applications?
You cannot simply sprinkle AI on top of a broken system; you have to do the heavy lifting of foundational work first to see any real results. We focused intensely on our core conversion and committed to a cloud-first strategy, which gave us a significant head start over peers who are still untangling legacy wires. This preparation allowed our engineering capabilities to improve exponentially, which is a total game-changer for a bank of our size and regional focus. We’ve seen the fruit of this labor through quick work on AI-powered prospecting and fraud capabilities, reaching production much faster than other similar-sized banks our partners are working with. It feels like we’ve finally unlocked a level of execution where we aren’t just talking about possibilities but actually seeing them live and breathing in our environment.
There is often a disconnect between signing a contract for an AI tool and actually seeing it generate value. How do you distinguish between merely identifying a use case and successfully bringing a tool into full production?
The market is currently saturated with people who talk about use cases and sign contracts, but the real challenge is sticking to the process until you realize tangible benefits. According to a survey of 73 banks by D.A. Davidson, about 42% of banks are providing individual tools, and 35% have implemented quantifiable cases, but many struggle to cross the finish line. We don’t necessarily aim to be the market leader in the raw number of use cases; instead, we pride ourselves on our ability to execute and see projects through to fruition. It requires a disciplined approach to ensure that a tool isn’t just a shiny toy but a “force multiplier” that assists our employees in their daily tasks. Being able to look at a completed project and see internal efficiencies replacing low-quality, repeatable work is where the true competitive advantage lies.
We have seen massive companies like Uber burn through their entire multi-year AI coding budgets in just a few months. How do you maintain financial discipline while experimenting with such expensive, resource-heavy technology?
Financial discipline is at the heart of our strategy, especially when you look at our $31.9 million technology and equipment expenses from the first quarter, which rose 7% year-over-year. We are incredibly “token-conscious,” monitoring the actual usage of AI units to ensure we aren’t just “tokenmaxxing” or wasting resources for the sake of a trend. Unlike some firms that gave every employee open access and are now pulling back due to spiraling costs, we track our spending monthly and evaluate the “why” behind every single interaction. We aim to be responsible stewards of our capital, ensuring that every dollar spent on software licensing or data processing fees translates into a clear return on investment. It’s about being deliberate from the get-go rather than reacting to a massive, unexpected bill four months down the road.
How do you manage the internal rollout of these tools across a workforce of thousands to ensure that power users have what they need without exposing the bank to unnecessary risk or cost?
We’ve implemented a very specific tiered structure to manage access for our roughly 3,600 employees to prevent the “Wild West” of AI adoption. We created five distinct tiers, with two specialized groups dedicated solely to engineering, quality assurance, and model risk management. Only about 80 employees—a cross-functional group of “power users” in diverse roles—have open access to the most advanced tools to build custom agents. Most tasks can be handled by pre-built agents, and we’ve found that unleashing custom access to everyone is a recipe for an expensive disaster. This tiered approach allows us to maintain a continuous focus on ROI, ensuring our credit and sales teams get exponential value without the chaos of unmonitored experimentation.
Banking is inherently a relationship business. How do you integrate AI into customer prospecting and lead generation without making the process feel “creepy” or losing that personal touch?
This is a journey that requires a deep understanding that AI is a tool, not a replacement for human connection or intuition. In a 100-year-old bank like ours, we’ve learned that the behaviors around generating new relationships aren’t easily codified; AI won’t naturally solve the nuance of a face-to-face meeting or a long-term trust bond. We use it to enhance our prospecting and help bankers develop leads, but we are very careful to work in a way that remains respectful and “not creepy” to our clients. You can’t just buy a tool and expect customers to start kicking down your door; you still have to operate within a very people-centric model. The goal is to give our bankers better data and insights so they can be more present and effective when they are actually building that relationship.
What is your forecast for the future of midsize banking as AI continues to evolve?
I believe we will see a dramatic consolidation of tech stacks where specialized, niche vendors will struggle to compete with a bank’s internal ability to build custom AI agents. Within the next few years, the “force multiplier” effect of AI will allow midsize banks to operate with the technical sophistication of a global giant but with the localized, personal touch that has always been our hallmark. We will move away from tracking use cases as a metric of success and instead focus on the total integration of AI into every credit and sales decision. The banks that thrive will be those that stayed “token-conscious” and foundational-first, ensuring their human bankers are empowered rather than replaced. It is an era where execution and the ability to bring tools to production will be the only metrics that truly matter.
