AI Can Wait. Connecting Your Data Can't.

Here's the thing about AI in banking that nobody wants to say out loud: Generative AI is built for maybe 5% of what your institution actually does — the unique, unpredictable work that requires creative problem-solving. The other 95%, like cash reconciliation, compliance reporting and loan processing? That's repeatable work that needs efficiency and consistency, not creativity.

Yet the entire industry is being told to rush toward AI or get left behind. Board members read headlines about ChatGPT and ask their CEOs what they're doing about it. The result is a lot of expensive technology theater while the actual problem — fragmented data — goes ignored and unsolved.

According to recent research from BCG, only 25% of financial institutions have successfully integrated advanced AI capabilities into their strategic playbooks. The rest aren't failing because they picked the wrong algorithm. They're failing because their data is scattered across dozens of databases, with people even in the same department storing information in different places.

It’s a valuable lesson. You can't build intelligence on top of chaos.

Banking’s silent saboteur

The real challenge isn't adopting the latest generative AI platform. It's that financial institutions are juggling dozens or hundreds of vendor partnerships while trying to integrate loan origination systems, core banking platforms, payment processors, and dozens of other specialized tools. Nearly 80% of that data is unstructured, sitting in silos that don't talk to each other.

The operational cost is eye-opening once you know where to look. Every night, staff spend hours after closing on manual reconciliation. Customers repeat themselves across channels because nobody has the full picture. Lending officers have the lending data. Tellers have transaction data. Marketing has their piece. The contact center records complaints. Compliance has a completely separate universe. None of it connects.

When someone walks into a branch for a mortgage consultation, the staff member should instantly see their complete digital banking history, transaction patterns, and relationship depth. Instead, too often they're apologizing for not having access to information the customer assumes they already know.

Nobody likes to repeat themselves. Ask them to “say that one more time” too many times, and they’ll be running for the exit.

Why AI isn't the answer (yet)

AI does have a role. It excels at the edge cases — summarizing long-form content, detecting anomalies, suggesting potential actions. Machine learning models can analyze millions of transactions to spot suspicious patterns and dramatically reduce false positives. First National Bank of Texas built a buy-now-pay-later solution that processes over 300 applications weekly. A customer standing in the TV aisle can apply for a loan and have the funds in their account before they reach the checkout counter. That's the kind of experience that keeps customers from defecting to fintechs. But none of it works without connected data underneath.

But when a teller asks "what's this customer's balance?" you don't want an AI model that might hallucinate a different answer each time. You want a system that delivers the same accurate answer, instantly, every time. That's not an AI problem. That's a data problem.

Many financial institutions will tell you their biggest barrier to AI isn't the technology — it's fragmented data, which in banking translates into missed fraud signals, slow approvals and inconsistent customer experiences. The algorithm isn't the issue. The foundation is.

Rule-based process automation, by contrast, delivers three to 10 times ROI in the first year. Why? Because it addresses the 95% of work that needs predictability, not creativity. An automation script can send notifications for dormant loans, reconcile data between systems or process batch statements with the same result every time, without hallucinations or surprises.

The approach that actually works combines both: AI identifies anomalies and opportunities, automation acts on them within defined parameters and humans finalize anything irreversible. That orchestration is how Arizona's Hughes Federal Credit Union cut loan processing from three weeks to two hours.

The path forward: Modernize. Connect. Activate.

Rather than chasing the latest AI tools to satisfy board expectations, institutions need a framework that addresses the real problem in three parts: modernizing operations, connecting enterprise systems and activating data to gain real intelligence.

Modernization means reimagining workflows, not just digitizing them. What if instead of turning paper forms into PDFs, institutions created conversational applications that prepopulate with known member data and route automatically to the right staff? What if statements weren't just online, but built into personalized dashboards that highlight opportunities and suggest relevant products?

American Heritage Credit Union took this approach. They didn't just digitize their loan process — they reimagined it with real-time decision-making, streamlined processing, and redefined workflows that actually made sense. The result was a 200% increase in indirect loan approvals. Members who previously waited about a week to open accounts now complete the process in under three minutes, with staff never having to manually input ID documents into the system.

Connection creates operational coherence. Branch staff can instantly access digital banking activity during mortgage consultations. Loan officers see complete relationship history regardless of the channel from which the application originated. Contact center reps understand recent interactions across all touchpoints instead of chasing down answers they should already have at their fingertips.

Real connection — in the form of technological integration but also in the relationships banks try to build with customers — requires moving beyond point-to-point integrations and into unified data flows, in which manual reconciliation gives way to straight-through processing. People First Federal Credit Union connected previously siloed data across their operations and saw 180 additional loans and 130 new deposit products in just six months.

Activation transforms data movement into customer intelligence. A customer's home equity application, investment inquiry and college planning discussion aren't three separate events — they represent a holistic financial strategy that requires equally holistic insight.

This is the "right data, right person, right time" capability everyone talks about. Personalizing recommendations based on life stage and transaction patterns. Meeting customers where they are, whether they prefer digital channels or in-person conversations for complex decisions. TruMark Financial Credit Union got activation right and generated 11,000 mortgage referrals and $167 million in new mortgage revenue.

The competitive advantage institutions are chasing isn't going to come from the most sophisticated AI algorithms. It's going to come from data that drives consistent, personalized service across every channel. Data that enables staff to have meaningful conversations. Data that converts every banking interaction — routine or complex — into an opportunity to build trust.

The result is deeper customer relationships, sustainable revenue growth, operational efficiency that transforms cost centers into profit drivers and competitive advantage built on data intelligence instead of AI theater.

Gordon Flammer is SVP of data activation, Kinective