The End of the "Buy" Era: How Valley Bank is Reshaping Midsize Banking Through Agentic AI

For decades, the midsize banking sector has operated under a rigid, predictable playbook: identify a operational friction point, scan the market for a third-party vendor, and purchase a purpose-built software solution. This "buy" culture defined the infrastructure of regional financial institutions, creating a patchwork of niche products that solved immediate problems but often left banks saddled with siloed data and ballooning licensing costs.

However, the rapid emergence of agentic artificial intelligence is shattering this conventional wisdom. Valley Bank, the $64 billion-asset institution headquartered in Morristown, New Jersey, stands at the vanguard of this shift. According to Russ Barrett, the bank’s Chief Operating Officer, the rise of sophisticated AI agents is fundamentally altering the build-versus-buy calculus, enabling regional banks to transition from passive software consumers to active engineering architects.

The Paradigm Shift: From Vendor Dependence to Internal Agility

Historically, midsize banks have relied on a "slew of niche products" to handle specific workflows—everything from loan origination compliance to customer service ticketing. These solutions were effective in their era, but they often lacked the flexibility to adapt to modern, data-intensive demands.

"We are already replacing three external contracts with our ability to leverage AI," Barrett revealed in a recent interview. These aren’t minor operational tweaks; they represent the displacement of legacy software solutions that were previously deemed essential.

For the largest "too-big-to-fail" banks, an engineering-first mindset has long been standard. They have the capital to recruit massive software teams and build proprietary stacks from the ground up. For midsize players, however, this has traditionally been an impossible luxury. Barrett argues that the barrier to entry is evaporating. "Our ability to exponentially improve our engineering is totally a game-changer for a bank like Valley," he noted.

By leveraging generative AI and agentic workflows, the bank is discovering that it can recreate—and often improve upon—vendor-provided functionality in-house, significantly reducing long-term reliance on third-party licensing fees.

A Chronology of Strategic Preparation

Valley Bank’s current success with AI is not a byproduct of a sudden technological pivot; rather, it is the result of a multi-year foundational effort. Barrett attributes the bank’s head start to the completion of two massive, high-stakes infrastructure projects: a total core conversion and the implementation of a cloud-first strategy.

The Foundation (2020–2023)

• Core Conversion: Moving the bank’s primary ledger to a modern, cloud-native architecture was the essential precursor to AI readiness. Without a centralized, clean, and accessible data environment, AI models would have been starved of the high-quality input needed to drive meaningful insights.
• Cloud Migration: By shifting to a cloud-first model, Valley ensured that its data processing power could scale elastically. This allows the bank to deploy AI models without the latency and hardware constraints that frequently plague traditional, on-premises banking infrastructures.

The Execution Phase (2024–Present)

• Production Velocity: With the infrastructure in place, Valley began collaborating with partners on AI-powered prospecting and fraud detection. Unlike many peer institutions that remain stuck in the "proof-of-concept" purgatory, Valley has reached the production phase with these tools ahead of the industry curve.
• Internal Refinement: The bank is now moving beyond pilot testing, shifting its focus toward "token-conscious" scaling, where the cost and utility of every AI interaction are audited to ensure a direct contribution to ROI.

Supporting Data: The State of the Industry

The shift occurring at Valley Bank is reflective of a broader trend within the financial sector. According to a recent survey of 73 banks of varying asset sizes conducted by D.A. Davidson, the emergence of tangible, production-ready AI use cases has accelerated sharply over the last few months.

The survey data reveals a tiered level of maturity across the banking landscape:

• 42% of respondents are currently providing individual, general-purpose AI tools to their staff (e.g., LLM chatbots for drafting emails or summarizing documents).
• 35% of respondents have moved beyond basic tools and have successfully implemented quantifiable, high-value use cases that directly impact the bottom line.

While the number of use cases is one metric, Barrett suggests it is a misleading one. "A lot of people talk about the number of use cases," he remarked. "It’s a little bit more challenging to be able to sit there and stick to it to see exactly how to get a tool into production and realize benefits."

Valley Bank’s internal metrics track not just the volume of AI deployments, but the "time-to-value"—the duration between identifying a problem and observing a measurable improvement in operational efficiency or customer satisfaction.

Financial Stewardship: The "Tokenmaxxing" Problem

The explosion of AI has led to a cautionary tale in other sectors, with firms like Uber burning through entire multi-year AI budgets in mere months. This phenomenon—often termed "tokenmaxxing"—refers to the reckless, indiscriminate use of AI models without regard for the underlying costs of compute tokens.

Valley Bank has adopted a strictly disciplined financial approach. Executives track AI spending on a monthly basis, and the bank is highly protective of its specific investment figures. However, public disclosures offer a window into the costs: in the first quarter of this year, the bank’s technology, furniture, and equipment expenses rose approximately 7% year-over-year. This $31.9 million expenditure was largely driven by the expected increases in data processing fees and software licensing, costs the bank is actively working to offset by replacing legacy vendors.

To prevent budget bloat, Valley has implemented a stratified access model:

1. Tiered Access: Employees are divided into five distinct tiers based on their role and technical proficiency.
2. The "Power User" Cohort: Roughly 80 of the bank’s 3,600 employees are designated as "power users." This cross-functional group receives open access to custom agentic tools, allowing them to iterate and innovate.
3. Specialized Engineering Tiers: Two tiers are reserved specifically for engineering, quality assurance, and model risk management, focusing on specialized, high-governance AI environments.

By restricting access to high-cost, custom-built agents, the bank avoids the runaway costs associated with universal access, ensuring that expensive compute resources are reserved for high-ROI activities.

Implications: The Human-Centric Future of Banking

Perhaps the most critical lesson Valley Bank has learned during its AI journey is that technology cannot replace the fundamental "people-centric" model of banking. When testing AI-powered prospecting tools, the bank discovered that the nuances of relationship building—the complex, unspoken social signals involved in generating new business—cannot be fully codified by current AI models.

"This is a journey," Barrett stated. "You’re not 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."

Internal Efficiencies vs. Relationship Building

Valley views AI through two distinct lenses:

• The Force Multiplier: Using AI to handle low-quality, repetitive tasks. This frees up human employees to focus on higher-order work, essentially acting as an assistant that boosts the productivity of existing staff.
• The Relationship Enhancer: Using AI to assist bankers in identifying, reaching out to, and nurturing clients in a "not-creepy way." The goal is to use data-driven insights to make the human banker more effective, not to replace the banker with a chatbot.

Conclusion: The Path Forward

For Valley Bank, the return on investment for AI is not strictly defined by cost-cutting. While the reduction in vendor contracts is a clear financial win, the true value lies in the bank’s increased agility. By moving away from the "buy" model, Valley is developing an internal muscle memory that will serve it well as AI technology continues to evolve.

The message to the midsize banking sector is clear: the era of relying on third-party black-box solutions to solve every problem is ending. The banks that succeed in the next decade will be those that treat their infrastructure as a core competency rather than a utility. By remaining "token-conscious," maintaining a rigorous focus on production-ready outcomes, and keeping a human-centric approach to customer relationships, Valley Bank is proving that a regional institution can compete—and lead—in an AI-dominated landscape.