For midsize banks, buying from vendors has been the norm. But the rise of agentic artificial intelligence has upended the build-versus-buy calculus, Valley Bank’s chief operating officer said.

Midsize banks typically have a slew of niche products or purpose-built solutions they’ve acquired through vendors that solved a particular problem the bank had at a certain point in time, said Russ Barrett, the $64 billion-asset bank’s COO. But as banks employ AI to a greater degree, that’s changing.

“We are already replacing three external contracts with our ability to leverage AI,” Barrett said in a recent interview. Those were software solutions focused on specific workflows, he said.

While the largest banks already have robust engineering capabilities and a build rather than buy mindset, “our ability to exponentially improve our engineering is totally a game-changer for a bank like Valley,” he said.

Completing foundational work such as the bank's core conversion and seeing through a cloud-first strategy has given the Morristown, New Jersey-based bank a head start in making strides with AI, Barrett contended.

He pointed to the bank’s quick work with other companies on AI-powered prospecting and fraud capabilities, having reached the production phase sooner than similarly sized banks those partners are also working with.

Barrett made a distinction between the ability to identify a use, sign a vendor contract and begin working, and the ability to get a tool into production and realize benefits. 

“A lot of people talk about the first, and it's a little bit more challenging to be able to sit there and stick to it to see exactly how to get more to the latter,” he said. “We probably are not market-leading in the number of use cases, but we do feel our ability to execute and see it to fruition is definitely something.”

Based on its survey of 73 banks of various asset sizes, D.A. Davidson said Wednesday “the emergence of tangible use cases over the past few months is notable.” About 42% of bank respondents are providing individual user tools, and an additional 35% have identified and implemented quantifiable use cases.

Valley is using AI to realize internal efficiencies related to low-quality, repeatable work and as a “force multiplier” to assist employees, he said. But the bank also sees AI as a customer relationship-bolstering tool.

That involves identifying and working “in what I would call a not-creepy way to leverage AI to be a better relationship bank,” Barrett said, as well as employing AI to enhance lead generation and prospecting, to help bankers develop new client relationships.

Growing AI use has led to skyrocketing costs for some firms, such as ride share company Uber, which burned through its entire 2026 AI coding budget in four months.

Valley executives are comfortable with the bank’s level of spending on AI, which is tracked monthly, Barrett said. The lender declined to say how much it’s invested in its AI strategy to date. 

In the first quarter, the bank’s technology, furniture and equipment expenses rose about 7% year over year, to $31.9 million, largely driven by increases in data processing fees and software licensing costs. 

Valley aims to be responsible with its AI use, and is “very token-conscious,” Barrett said. 

Tokens, used as a unit of measure for AI use, have become a method for pricing services, although they only account for a portion of AI spend. That’s given rise to the term “tokenmaxxing,” which refers to maximizing AI use.

Barrett contrasted Valley’s deliberate approach from the get-go – to monitor actual token use and the “why” behind it – with that of companies that gave every employee access to pricey tools and are now pulling back.

Valley created five tiers of employees with varying levels of access to AI tools; two of those tiers are specifically for employees in engineering, quality assurance and model risk management, and involve specialized AI tools, Barrett said. 

There are plenty of tasks that can be aided by already-built AI agents, rather than unleashing access for any employee to build custom agents, and the latter becomes expensive quickly, he said.

About 80 of Valley’s roughly 3,600 employees are in the tier that has open access, and that’s a cross-functional group of “power users” in diverse roles across the bank, Barrett said.

“There is a continuous focus on [return on investment], and the ROI doesn't obviously always have to be cost saves,” he said. “It’s a formula that allows us to really understand, how do we exponentially add more value to our credit people, to our salespeople, than what we would have today?”

Experimentation has also offered the bank some key lessons. With its AI-powered prospecting capability, the 100-year-old bank has learned “the behaviors around how you generate new relationships are, today, not codified in a way that AI is going to naturally solve,” he said. 

“This is a journey, and it is not, you're going to go buy a tool, and all of a sudden your door is going to be kicked down with all the customers you're able to generate, because you're still operating with a very people-centric model,” he said.