AI’s Delicate Banking Balancing Act

Banks face AI disruption after years of recovery; efficiency gains and higher returns are in sight, but adoption is slow, costly—and risky.

Banks around the world spent years on the defensive after the 2008 global financial crisis, repairing balance sheets, absorbing a flood of new regulations, struggling with record-low interest rates, and ceding market share to private credit and other unregulated providers. The past few years brought respite as central banks tightened policy, bolstering interest income, and economies accelerated out of the pandemic.

Now comes disruption from AI.

Any C-suite in finance can see AI’s potential benefits in automating labor-intensive paperwork or spiffing up customer interaction. More than 90% of banks are “already in the AI journey,” says Sameer Gupta, Americas financial-services AI leader at EY.

The rewards could be meaningful. S&P Global estimates that rated banks’ average return on equity could rise to 14% from 12% over the next three to five years, says Miriam Fernandez, the agency’s Madrid-based lead researcher on AI adoption.

The trip won’t be easy, though. Experiencing the wow factor from querying ChatGPT or “vibe coding” with Claude is one thing. Standardizing those leaps across large, data-dense organizations is quite another; and it’s both laborious and expensive. “Diffusion takes longer than people think,” says Alexandra Mousavizadeh, co-founder and co-CEO of Evident Insights, a London consultancy that tracks AI in banking. “You’re changing habits, not just flipping a switch.”

The AI revolution is often frustrating. Fernandez estimates that nearly half of all initiatives within banks fail. “Not every dollar or euro invested results in a solution, and not every solution can scale to where return on investment becomes tangible,” she says.

Technology could become increasingly dangerous for an industry that is the world economy’s circulatory system. As ChatGPT-style generative AI evolves into agentic AI models, these models will likely make more decisions on their own.

In April, US Treasury Secretary Scott Bessent and Federal Reserve Chairman Jerome Powell brought in top US bankers to warn about the cybersecurity risks associated with Mythos, the latest innovation from Anthropic, the creator of Claude. A company statement noted that Mythos had uncovered flaws in existing computer operating systems that “have in some cases survived decades of human review and millions of automated security tests”—a scary prospect in the wrong hands. Anthropic agreed to restrict the model for now to a handful of corporate clients, in “an urgent attempt to put these capabilities to work for defensive purposes.”

Bankers cannot rely on Silicon Valley or their regulators to protect them from every AI innovation, says Eric Alter, who recently retired as an AI engagement leader at global professional-services firm Marsh in the UK. “The ‘tech bros’ are still very much in their hype cycle,” he says. “If bankers aren’t careful, we’re screwed.”

AI’s Initial Boring Benefits

As “tech-bro” wizardry meets risk-obsessed, regulated commercial banking, the first benefits are likely to be more boring than dramatic. The outside world may not see much at all for the next few years, or longer. More than 85% of current AI use cases are internal, Mousavizadeh estimates.

That does not mean AI is irrelevant to the bottom line. An AI agent might cut onboarding time for large new clients from “six months to six weeks,” for instance. “The agent can reach into databases, go back and forth checking documents for know-your-customer,” she says. Investment bankers scrambling for analytical and legal documentation on a proposed merger might similarly “take out a lot of hours.”

Shahmir Khaliq, head of services at Citi in New York, adds more-efficient treasury management and custody operations to AI’s to-do list—major priorities in the banking world that won’t grab headlines outside of it.

“We’ll see efficiencies first before we see a lot of visible innovation,” echoes JoAnn Stonier, a former chief data officer at Mastercard who now teaches at Carnegie Mellon University in Pittsburgh.

Focusing on internal processes also helps mitigate security risks. “There is a box, and we operate within that box,” Khaliq says.

Risks Sneak In The ‘Back Door’

Moving out of the box, bankers must worry about the other guy’s agent, not just their own. “More tech-savvy customers will want their own AI tool interacting with the bank’s,” EY’s Gupta says. That could compound headaches in the race to establish AI standards, which are wide open globally.

Deploying AI across payment systems that connect hundreds of banks and millions of merchants is even more challenging, Stonier adds. “Agents have to work across an ecosystem where Mastercard is using one large language model and the counterparty another,” she says. “We don’t have the protocols yet for how they metaphorically shake hands.”

Nor are the firewalls necessarily sufficient to prevent a rogue AI agent from infecting or coopting other systems. “Risk is coming in through the back door, with vendors’ agents liaising with each other,” Evident’s Mousavizadeh says.

Bankers and AI pioneers operate at fundamentally different speeds. “Any financial institution works on a three- to five-year plan, while the tech horizon is six to 12 months,” says Alter. “By the time a tool is deployed, it’s obsolete.”

Regulators and legislators often lag behind both technology and industry, reacting forcefully only after a crisis—which, given AI’s potential power, may be too late. “It’s very difficult for the laws to keep up,” Stonier says. “It’s up to organizations to retain the trust of customers so they can stay in business.”

Alter advises, “You’re almost better off waiting and letting other people make the mistakes.”

Racing To Stay Ahead

Few bank bosses can afford to take that advice and risk being left behind. “Banks that secure the benefits of AI—including across costs and revenues—could find themselves with enduring advantages over competitors,” states an S&P Global special report from last October. “We expect rated entities’ financial and competitive positions to diverge within the next three to five years.”

One key to victory in the AI-innovation race is the groundwork banks should have laid years ago, as paper records became digital and digital data migrated to the cloud. AI models will be only as good as the terabytes of data they are gobbling up. “Data readiness, working with data sets that are clean and not duplicated, is a source of competitive advantage now,” S&P’s Fernandez says.

Another is accessing the limited pool of AI professionals and harnessing their expertise. “Talent is the key factor to solve for,” Gupta says. “The creators and providers of the technology are also creating programs for people to learn it.”

US banks have a natural advantage here, given the proximity of so many creators and providers, Mousavizadeh argues. Seven of the top 10 names in Evident’s latest AI Banking Index are headquartered in North America, led by JPMorgan Chase, Capital One, and Royal Bank of Canada. “European banks are one step behind, without the same access to an AI startup ecosystem,” she says. “The US has a lot of open doors for talent to move back and forth.”

According to Evident, just three US banks—JPMorgan Chase, Capital One, and Bank of America—account for 75% of the industry’s AI-related patents.

Some European companies seem to be doing more with less by devising game-changing, customer-facing initiatives. Fernandez points to the groundbreaking DealSync platform from Italy’s UniCredit, which identifies thousands of merger or acquisition opportunities for midsize companies across Austria and Germany, as well as Italy.

Germany’s Commerzbank was ahead of the pack in rolling out an AI-generated customer service avatar named Ava a year ago. It is fielding more than 30,000 inquiries a month and resolving three-quarters of them. “She blends empathy with engineering precision,” according to a company release.

Dutch giant ING stands out for using AI in banking call centers, a frequent source of customer frustration everywhere, Alter says. ING’s agent handles routine inquiries smoothly but is programmed to connect to a human for sensitive life events.

ING’s next focus is enabling mortgage applications solely through interaction with an AI agent, says Chief Operations Officer Marnix van Stiphout. “It is aiming for a rollout this year.”

Speed Vs. Trust

It’s become a cliché to say that a new technology’s effects are overestimated in the short term and underestimated in the long term. That truism stems from the internet, which revolutionized banking and the rest of the world a quarter century ago, though not as quickly as many credulous investors and managers assumed. “Internet diffusion took about 10 years,” Mousavizadeh reflects. “It’s never a plug-and-play process.” She expects AI to move roughly twice as fast—which still means five years of diffusion.

A less-noted lesson of the world’s going online is that banks are among the most stable institutions, if only because societies can’t live without them. Most of today’s global tech giants barely existed, if at all, at the turn of the 21st century. The big banks are still, by and large, the big banks—trusted despite or because of their slow-moving, change-suspicious nature and the perception that they are too big to fail.

Trust could be a more precious asset than ever in an age increasingly defined by distrust of government, information, technology, and everything else. Internet pioneers were heroes, wrapped in utopian visions, for decades before their halos corroded. AI is already surrounded by prophets of doom, even before it has achieved anything practical. One big mistake with it, optimists and pessimists agree, could cost a franchise.

None of that means bankers can stand by and wait to see how things work o before diving into AI. “People need time to get adjusted,” Mousavizadeh says. “It’s a bit of hand-to-hand combat.”

It means proceeding with caution and patience. S&P’s Fernandez says, “Moving so fast, it’s very hard to get the right balance of risk and innovation.”