Big Banks Are Using AI to "Out-Code" Vendors. Here's How Community Banks Can Still Compete.

In a recent American Banker article, Fifth Third Bank executive Jay Plum made a provocative claim: artificial intelligence will allow large banks to "out-code" their technology vendors. Instead of waiting on Fiserv or Jack Henry to ship features, banks with internal AI teams will build what they need themselves, faster and cheaper. Plum described how Fifth Third is already using AI to run ongoing "mini reviews" of its loan portfolio and risk exposure, replacing the traditional annual scramble of pulling data from a dozen systems to prepare for regulatory exams. The bank scans portfolios continuously, flags emerging concentrations early, and maintains exam-ready documentation as a byproduct of daily operations rather than a quarterly fire drill.

But the underlying insight is worth paying attention to, because the operational outcomes Plum is describing are not exclusive to large banks. The delivery model just has to look different.

What Fifth Third Is Really Solving

Strip away the "out-code" language and Plum is describing three operational improvements that every bank CFO and COO would recognize as valuable.

1. Continuous exam readiness instead of periodic scrambles. Rather than spending weeks before an exam pulling data from core, LOS, GL, and compliance systems, the bank maintains a continuously updated, auditable record. When examiners ask a question, the answer is already assembled.

2. Earlier risk detection through ongoing portfolio analysis. Instead of discovering loan concentrations or documentation gaps during annual reviews, AI-powered monitoring surfaces them in near real time. Issues get addressed before they become findings.

3. Less dependence on vendor release cycles for critical workflows. When a bank can build and adapt its own tools, it stops waiting 12-18 months for a vendor to prioritize a feature request. It controls its own operational roadmap.

These are outcomes, not technology features. And that distinction matters.

The Scale Advantage Is Real

The honest assessment: replicating Fifth Third's approach in-house is not realistic for most community institutions. To "out-code" vendors, a bank needs a modern cloud-ready data environment that unifies information across core, LOS, document repositories, and compliance systems. It needs an internal engineering team that understands both banking operations and AI tooling. And it needs the ability to operationalize AI models, which means embedding them into daily workflows, monitoring performance, and documenting everything for model risk management and regulatory scrutiny.

For small to mid-size community banks ($500M - $10Bn), dedicating that level of budget and specialized talent to internal AI development is rarely justifiable. The math doesn't work. A single senior data engineer costs more than most community bank technology budgets can absorb, and that's before infrastructure, tooling, and the compliance overhead of running production AI systems in a regulated environment.

This is the "cost vs. capability" trap that keeps community banks stuck between enterprise software they can't fully implement and manual processes they can't scale.

Same Expectations, Fewer Resources

Regulatory requirements scale with asset size. A $600 million community bank isn't filing the same reports or facing the same supervisory cadence as a $50 billion regional. But the standard of evidence doesn't scale down with the balance sheet. Examiners still expect accurate data, defensible processes, and clear traceability - regardless of how many people are on the ops team producing it. That asymmetry is the core problem: smaller banks carry a proportionally heavier compliance burden relative to their resources, even though the absolute reporting load is lighter.

The OCC's 2025 Request for Information on community bank digitalization confirmed what most operations leaders already know.

• Banks described compliance teams spending weeks preparing for exams because data lives in five or six different systems.

• Loan review teams spending half their time extracting information from PDFs before starting the actual analysis.

• New technology investments underperforming because the underlying data is fragmented, unstructured, or trapped in legacy formats.

In that context, Fifth Third's "mini review" approach is less a technology story and more a process story. Continuous monitoring, automated exception handling, and built-in documentation relieve exam pressure and reduce operational risk.

Five Things Community Banks Can Do Right Now

Before getting into operating models and partnerships, it's worth acknowledging that community banks are not helpless here. There are practical, concrete steps that any institution can take to start closing the operational gap, regardless of budget or technology maturity.

None of these require a seven-figure technology investment. They require operational clarity and a willingness to look honestly at how work gets done today.

From "Out-Coding" to "Out-Executing"

The goal shifts from building an internal AI stack to achieving the same operational outcomes through a combination of smarter process design, selective technology adoption, and the right partnerships. There are several ways to get there. Some banks renegotiate with their core provider to unlock integration capabilities they're already paying for. Others invest in lightweight automation tools their existing staff can manage. And some look for partners who can take ownership of specific workflows and deliver clean, validated, audit-ready data as an output - freeing internal teams to focus on higher-value work. The common thread is that none of these paths require hiring a team of AI engineers. They require clarity on which operational problems actually need solving and an honest assessment of whether the bank has the internal capacity to solve them alone.

This model looks different from traditional software procurement or BPO staffing in a few important ways.

• The bank buys an outcome, not a tool. Rather than licensing software and hiring staff to run it, the bank pays for a defined result at a predictable cost. Loan files are complete and validated. KYC documentation is organized and defensible. Reconciliation happens same-day with exceptions routed automatically.

• The data infrastructure is handled. Document intake, classification, extraction, cross-validation, and lineage tracking all happen within the managed workflow. The bank doesn't need to build a cloud-native data pipeline or maintain AI models internally.

• Automation without blind spots. Any automation approach. Whether built internally, purchased, or delivered through a partner - human checkpoints are needed for the exceptions that automated systems can't resolve confidently. The percentage varies by workflow, but the principle holds: full automation without oversight creates a different kind of risk.

• Audit trails exist by default. Every data point, every transformation, every human review decision is documented with full lineage. When examiners ask "show me your process," the answer already exists.

What Does This Look Like in Practice?

To make the concept concrete, consider how managed data operations can address the same pain points Fifth Third is solving with internal AI teams.

These are not theoretical capabilities. They mirror exactly what Fifth Third is building internally, delivered as a service rather than as a large-scale engineering project.

Competing Without Writing Code

Jay Plum's argument is that AI shifts the balance of power between banks and vendors. Banks with strong internal teams can build and adapt faster than vendor roadmaps allow.

For community banks, the same logic applies with a different mechanism. The competitive advantage doesn't come from out-building the largest players on technology. It comes from being more deliberate about which problems to solve first, more honest about internal capacity, and more open to operating models that didn't exist five years ago.

The options are broader than the old binary of "buy enterprise software" or "hire more people." Banks can push their core providers harder on integration and data access. They can adopt targeted automation for specific high-volume workflows. They can bring in partners for the operational work that doesn't justify a permanent internal team. And they can do all three in combination, starting small and scaling based on what actually works.

In a market where large banks are racing to out-code their vendors, community banks have a viable counter-strategy: focus on out-executing on the things that actually matter. Exam readiness. Loan cycle time. Data accuracy. Operational cost control. The competitive advantage for community banks was never going to come from building a bigger technology stack. It comes from staying focused on what they do best - serving their communities - while making sure the operational backbone keeps up.