How AI is “flipping the pyramid” of the banking business model: Sibos 2025

“AI is going to impact every aspect of banking as we know it,” said Shanker Ramamurthy, IBM Global Managing Partner for Banking and Financial Markets, speaking from the floor of the Messe Frankfurt at the Sibos conference this week.

In his talk on “Scaling AI within the banking enterprise,” Ramamurthy explored how forward-thinking financial institutions are moving from AI experimentation to execution. He homed in on “the three things that keep CEOs and CXOs awake at night”: growth and performance, cost and efficiency, and regulatory compliance posture and mitigating risk. AI will be central to all three, he explained.

Ramamurthy cited a recent IBM Institute for Business Value study, 2025 Global Outlook for Banking and Financial Markets, which highlighted the reality of return on equity (RoE) in today’s banking industry. “Banks in the aggregate globally do not achieve their cost of capital,” he said. In other words, the typical bank’s RoE is lower than the rate of return required to cover the risk of the investment.

That gap usually resolves itself in banking through mergers and acquisitions, Ramamurthy said. Thus, he explained, “we can state with confidence that over the next three years, you’re going to see a lot more M&A activity. In mature markets like the US, consolidation has been the order of the day already; we’re going to see acceleration.”

Another major theme in the banking world is rethinking cost and efficiency, he said. Cost-income ratios—the ratio of operating costs to operating income—“have been stubbornly stuck, despite banks doing a lot of work to address them.” He cited data from The 94% Core Banking Problem, a new IBM survey of CIOs: In the US, for example, the cost-income ratio is as high as 62%; in Europe, it’s around 54%; and in Latin America and Asia-Pacific, it’s lower still. “There’s lots of opportunity there in mature markets,” he said.

Risk mitigation and compliance issues, said Ramamurthy, “are an area where banks are investing disproportionately.” The reason, he explained, is obvious: increased regulation on a global level. “A multi-jurisdictional global bank might have 12,000 discrete regulations around the world to comply with,” he said.

Furthermore, generative AI (gen AI) is going to enable the creation of more software, leading to a larger cybersecurity attack surface, Ramamurthy said.

Not to mention that large-scale, fault-tolerant quantum computing is just around the corner, he said, noting IBM CEO Arvind Krishna’s projection that it would be available by 2029. One potential downside of this giant leap, Ramamurthy said, is that it can compound risk. “Bad actors are looking to steal encrypted data—social security numbers, for example—and sit on it, waiting for quantum to be available. On the other hand, quantum computing enables solving problems that traditional computers are not able to address effectively or in a timely manner, including optimization of prices, efficient frontier simulation, real-time risk management and more—thereby representing huge opportunities in the years ahead.”

For the past five decades, Ramamurthy said, “some two-thirds to three-fourths of investments in banking technology, people and processes have been focused on the middle and back office,” with the remainder focused on the front office: “customers, channels, ecosystems, platforms and partnerships.” This is no longer sustainable, he said, because it’s led to “complex, inflexible, expensive technology and processes supported by monolithic systems that make change really hard.”

That’s yesterday’s model, Ramamurthy said, adding that the banking model of tomorrow will be “radically reimagined” and “will flip the pyramid on its head.” Gen AI, hybrid cloud and eventually quantum computing will simplify middle- and back-office processes, freeing up banks to focus on the customer. “That transition from yesterday to tomorrow’s business model is not optional,” he said.

One peculiarity of AI technology, Ramamurthy said, is that it marks “the first time in my career I’ve seen other industries adopt a technology much more broadly and rapidly on an enterprise scale than banking.” Stands to reason, he said, given the unique risk and regulatory exposure financial institutions have to manage. Though change is coming, and fast. According to The 94% Core Banking Problem, over 90% plan to incorporate AI by 2028.

“This is an extraordinarily interesting time for the banking industry as a whole,” he said. “It’s not just business driving the technology … technology capability is actually informing business strategy.”

The biggest benefit financial institutions get from AI, especially gen AI, is in the software development life cycle [SDLC], where productivity might improve by 20%, Ramamurthy said. He cited a stunning real-life example: Bradesco, a large bank in Latin America and Brazil, uses gen AI to handle its 283,000 monthly customer queries, reducing wait times from 10 minutes to a few seconds.

IBM’s best case study is perhaps IBM itself. As Ramamurthy said, “We eat our own cooking.” Offering itself as “client zero,” IBM tests AI capabilities internally before rolling them out to clients, he explained. The results are impressive: over 86% of IT issues raised by IBM employees are handled through the gen AI tool “Ask IT”; meanwhile, 94% of HR queries are handled by “Ask HR.” All told, “this translates into about USD 4.5 billion of efficiency gains on an annual basis,” he said.

These efficiencies, he said, have allowed IBM employees to “operate two band levels higher” than before, “thereby upskilling the employees, making their jobs more fun, more productive and more valuable.”

Financial technology companies (fintechs) have been competing with banks for years. A more formidable threat, Ramamurthy said, is the so-called “techfin”—tech companies that enter the financial services arena, as opposed to startups, like Stripe, that originated as finance companies.

Apple, Google and Alibaba are all examples of techfins. Going forward, both fintechs and techfins will be “coming after some of the most profitable parts of the banking franchise,” Ramamurthy said. But when it comes to the AI transition, banking has a treasure hidden in plain sight that these competitors cannot replicate. “In this new world, the natural and fair advantage for a bank is the data,” he said.

Whereas most gen AI models have already harvested nearly all publicly available data, he said, banks have highly valuable proprietary data: the voluminous data surrounding customer transactions. “Within financial institutions, I would venture to say 95% or more of the data is yet to be mastered in the generative AI context,” Ramamurthy said.

For the banking industry to stay competitive in a rapidly evolving AI ecosystem, Ramamurthy stressed, seizing the moment is crucial. “The opportunity is here and now for financial institutions to accelerate toward that business model,” he said.