The banking sector’s adoption of AI, experts agree, is neither sudden nor complete. What’s unfolding instead is an evolution — a steady embedding of machine intelligence into human systems of judgment, compliance, and care. From lending to fraud detection, the new wave of intelligent automation is reshaping how financial institutions think, decide, and serve.
A New Intelligence at Work
For decades, banks have been known for their silos — separate departments that barely speak to each other. But now, artificial intelligence is forcing a conversation across those walls. Executives from Cornerstone Advisors, a U.S.-based consulting firm, say that AI has become less a “function” and more a force — one that must be woven into every operational layer of finance.
“Banks and credit unions have to stop thinking about AI as a function,” said Tony DeSanctis, senior director at Cornerstone. “We need to be thinking about AI being integrated into everything we do.”
This shift, DeSanctis and his colleagues argue, marks a fundamental reorganization of financial institutions — where data, decision-making, and customer engagement are being re-coded by intelligent systems. The industry’s leaders are now navigating a balance between human judgment and algorithmic precision, in areas ranging from loan approval to customer contact.
Lending: Balancing Human Judgment and Machine Logic
Artificial intelligence has already begun to prove its worth in lending, according to Daryl Jones, senior director at Cornerstone. Banks using AI in decision-making have reported “healthy increases in approvals,” he said, while maintaining strong performance.
But the shift isn’t without complications. Jones cautioned that human biases — the very traits AI was designed to transcend — can creep into systems through flawed training data.
“Aspects of loan decisions are transitioning to AI,” he said. “Institutions have to be careful how human lenders influence the software development process.”
The problem isn’t hypothetical. DeSanctis recalled one company that trained an AI hiring tool using résumés from its top male software engineers. The algorithm quickly learned to favor men, filtering out female graduates from historically women’s colleges.
“AI has zero conscience and zero ethics,” he said. “Banks, subject to fair-lending and compliance rules, have to anticipate potential problems.”
Yet, when properly built, these systems can spot refinancing opportunities, identify borrowers, and initiate early contact — tasks that once consumed entire teams of analysts.
From Fraud to Trust: The Risk Management Revolution
If lending defines the promise of AI in banking, risk management reveals its power. John Meyer, another Cornerstone managing director, says financial institutions are moving from “day-two detection” — catching fraud after it happens — to “in-the-moment interdiction.”
“We’ve had machine learning since the 1990s,” Meyer said. “What’s different now is that AI can evaluate behaviors in real time.”
Modern antifraud systems use subtle behavioral cues — how a user types, moves their cursor, or navigates an app — to distinguish genuine customers from impostors. When anomalies appear, the AI can instantly trigger an extra authentication layer, like a callback or multi-factor check.
“Fraud is always a math problem,” Meyer added. “Am I stopping more than I’m spending on fraud systems?”
Voice biometrics are also advancing verification. As Ryan Brogan of Cornerstone noted, AI models can now recognize the cadence and tone of a caller’s voice within seconds. Still, he warned that “deepfakes” of audio and video are making such defenses vulnerable.
“Voice biometrics won’t be sufficient on its own,” he said. “It’ll be part of a layered approach.”
The New Frontiers: Collections, Conversations, and the Agentic Wallet
Beyond lending and fraud, AI is quietly transforming the back office and customer touchpoints. Chatbots, document processors, and predictive collectors are becoming integral to the way banks operate.
Jones described new systems that analyze borrower behavior to optimize collection efforts:
“When a borrower’s called, what time of day are they picking up? Do they respond better to texts than to phone calls?”
In another corner of the office, automated assistants are cutting customer call times nearly in half. These “agent-assist” tools combine natural language processing and retrieval-augmented generation to help representatives resolve complex queries in seconds.
Even payments are entering an era of intelligent autonomy. DeSanctis envisions the rise of the “agentic wallet” — a system where an AI makes purchases and payments on behalf of consumers.
“Being ‘top of the agentic wallet’ is going to become critical,” he said. “You’ll want to make sure your cards get added into it.”
For back-office managers like Clio Silman, AI’s value lies in order. Her clients once struggled with operational chaos — “Word files on someone’s drive, PowerPoints on another.” Now, centralized AI tools not only organize documentation but continuously update it as procedures change. “It’s the project that keeps on giving,” she said.
As DeSanctis put it,
The question is no longer whether banks should use AI, but how deeply it should be trusted to think for them.
