AI in banking ‘not a silver bullet’

Where’s the ROI from AI in banking? This piece from Banking Dive reminds us that, despite years of near constant effort to deploy AI, the industry is still very much in the early days when it comes to real results.

Bradley Leimer, Founder & Principal at Leimer One Advisors said: „The real story isn’t that AI isn’t working, it’s that we’re still building the infrastructure, the operating models, and the capabilities to use it well. From my time working on AI initiatives at SMBC, the real unlock wasn’t the AI operating model, it was aligning people, process, partnerships, and the purpose around it. It was about consistently driving AI across the organization, providing space for experiments that can lead to solid integrations with existing data and systems. It was about rethinking existing workflows. That’s where true reinvention and transformation will take place.„

Awa Fabrice: „The patience piece is what’s hardest to measure. ROI on AI shows up in efficiency and internal workflows first, but the compounding effect comes later, when the infrastructure, data, and operating model are finally aligned. In accounting, we call that ‘intangible value.’ Hard to put in a quarterly report, but it’s what makes the next leap possible.„

Gaurav S. Chug, Chief Executive Officer of Synergy Resources LLC: „The “smart co-worker” framing feels right , AI today is mostly optimizing within existing systems so the ROI shows up as efficiency. The bigger shift likely comes when we rethink the infrastructure layer itself especially how trust and data move across systems and borders. That’s when it moves from incremental gains to real transformation.„

Michel Capuis, Partner SNGLR Group: „From my work with FS clients in Switzerland and DACH, I see the same: real ROI from AI comes slow. The bottleneck is rarely the model, it’s data, decision rights, and old workflows. Efficiency shows up first, but real revenue impact is a marathon, not a sprint.„

Ron Shevlin, Chief Research Officer at Cornerstone Advisors commented: „Value and ROI are 2 different things. This is the same argument that the business world had 40 years ago–„what’s the ROI of personal computers?” There is no ROI on infrastructure. The ROI comes from initiatives that use the infrastructure. AI=infrastructure. I guess we haven’t learned anything.”

Please find below the full article written by Banking Dive

Despite the buzz around artificial intelligence’s potential to transform banking, it’s no “silver bullet,” according to Mike Mayo, a top bank analyst from Wells Fargo.t.

“It is a long, expensive and risk-constrained transformation,” Mike Mayo wrote in a note to investors last week. “Indeed, we can’t find one extremely noteworthy new product or service from AI in these very early days (it’s been a little boring from a product standpoint).”

Actual payoffs will take time and the human in the loop isn’t disappearing soon, Mayo wrote Tuesday, following a bank and tech summit featuring input from JPMorgan Chase, Bank of America, Wells Fargo, Goldman Sachs, BNY, Amazon, Meta and a former Nvidia executive.

“The long-term upside,” Mayo wrote, “belongs to scale players that pair proprietary data with disciplined process redesign, with the added benefit that trust is a moat.”

Sean Viergutz, PwC’s banking and capital markets advisory leader, has seen a shift toward that process redesign mindset in the last six to 12 months, as AI understanding grows. 

“Those that have really grasped the power of it don’t use it as a tool,” Viergutz said during a recent interview. “They use it as a co-worker. They use it as a teammate.”

There’s more consideration of “how are we fundamentally changing [organizational] structures, process structures, ways of working, operating procedures to really harness and involve the digital workers part of this?” he said.

Goldman Sachs CEO David Solomon alluded to that in a shareholder letter, released Friday, mentioning the bank’s new operating model “propelled by AI.”

“It has become increasingly clear that our operating processes need to reflect the gains that will come from these transformational technologies,” Solomon wrote. 

Still, Viergutz said, “there’s a perception gap of, ‘Oh, AI can do absolutely everything,’ and that’s certainly not the case.”

While it can be “a step-level shift into being more efficient, being more accurate,” he said, “it’s not going to do the job of the expert worker that understands the systems, the technology and what has to happen in the business flow of things, or the product flow of things.”

Among other areas, banks are employing AI in engineering and code development, know-your-customer and anti-money laundering applications, and investment banking, where automation means credit analysts aren’t spending hours putting pitchbooks together, Viergutz said. 

Mayo cited an estimate that one-third of bank jobs or portions of jobs may eventually be better handled by AI.

“For money center banks, there is more scale and productivity across massive headcounts. For trust/asset servicing banks, there’s a greater focus on agentic AI for document-heavy, data intensive workflows. For regional banks, efforts are more for targeted efficiency (call centers, fraud, coding),” Mayo wrote.

The bank and tech summit Mayo referenced “reinforced our view that AI can help banks trend toward record efficiency. Banks are transitioning from pilots to production to profits. Yet, the question remains when AI dreams turn to dollars.”

ROI questions
Jamie Dimon, the CEO of JPMorgan, the biggest U.S. bank, told Bloomberg last October the bank’s annual AI-generated savings of $2 billion is about equal to what it spends yearly on the technology.

Other than JPMorgan, “almost no banks disclose savings,” Mayo wrote, adding he wonders if banks know the return on investment they’re currently getting.

It’s a common question right now, as the industry experiments with agentic AI, said Nina Owens, managing director and co-lead of financial services strategy at Publicis Sapient.

“We haven’t had 20 years of agents, or even five years of agents, to really capture what the typical returns are,” she said in a recent interview.

Keri Smith, Accenture’s global banking data and AI lead, expects to see firms redesigning job roles to support AI adoption over the next six to eight months, as companies realize what’s needed to see results. She called about 8% of financial services firms AI front-runners, meaning they’re reinventing workflows, have organizational alignment and are already seeing results.

Owens said several of her bank clients aim to revamp marketing operations with agentic AI, keeping the marketer or content manager in the loop while agents handle data-pulling or other rote work.

That’s become a big focus as banks seek to grow their businesses, she said, because agentic AI can quickly connect the data and identity dots so a bank can provide hyperpersonalized offers to customers. To get there, though, a lot of foundational work has to occur, related to data and modernizing systems, she noted.

Can’t be ‘AI-ed overnight’
At Bank of America, an AI catalyst group has Jeff Busconi, the bank’s head of strategy, and Hari Gopalkrishnan, BofA’s chief technology and information officer, working with 18 senior business leaders representing different areas of the bank, “to make sure that we’re driving AI into every single area,” said Dean Athanasia, the lender’s co-president, during an appearance at an investor conference this month.

Given the bank’s size, “it’s a way to get it done and to share and leverage and get every bang for every investment dollar,” he said.

BofA has said AI-powered virtual assistant Erica handles the work of about 11,000 people, and all 18,000 of the bank’s software developers use coding agents to optimize the development process, boosting productivity by about 20%. The bank has also rolled out Salesforce’s Agentforce, which enables the creation of AI agents to handle tasks, to 1,000 financial advisers, Athanasia said.

Executives at fintech Chime envision AI bolstering efficiency with compliance reviews and ensuring compliance with unfair, deceptive, or abusive acts and practices standards, CEO Chris Britt said at an investor conference this month.

“You still need some element of human involvement, but there’s so many ways to make these processes more and more efficient,” he said.

It takes time to get it right, though, Britt said, because the fintech operates in a highly regulated business requiring “deep interactions” with banks and regulators.

“It’s not something that can just be AI-ed overnight,” Britt said.

Indeed, cautious approaches to AI remain key for big banks, Mayo emphasized, and precision is more important than speed.

“AI failures at banks have greater downside than others since they erode a bank’s most important asset: trust,” Mayo wrote.