TLDR
- Most retail banks are still stuck in AI pilots, with only about 5% seeing real bottom-line impact.
- Leadership, not technology, is the biggest bottleneck holding banks back from scaling generative AI.
- AI-first banks could cut costs by up to 40% and boost profits by 30% if they move decisively.
Retail banks are talking loudly about generative AI. Much more quietly, most are still stuck in experimental mode.
That gap between ambition and impact is now one of the biggest strategic risks facing the industry, according to Juan Uribe, managing director and senior partner at Boston Consulting Group and its North American leader for retail banking.
“The reality is on a bunch of dimensions, we’re in the early innings,” Uribe said in an interview with The AI Innovator. “Many seem to be dabbling in AI.” But without deploying AI into production, he estimates that only 5% of companies are “actually getting meaningful bottom-line impact.”
That assessment is echoed in BCG’s latest Retail Banking Report, which Uribe co-authored. The report said AI could unlock more than $370 billion in additional annual profit industrywide through efficiency and growth, yet most banks are not moving quickly and decisively enough to benefit.
Banks often mistake activity for progress. “There’s an element of letting a thousand flowers bloom,” Uribe said. “If you ask the banks, ‘Are you doing anything in gen AI?’ They say ‘Yes, because we have all of these pilots. We’ve encouraged everyone to go and do something.’ But the reality is that it’s not really driving any real value or any real impact.”
BCG’s research shows that banks rate themselves highly on AI adoption, ranking behind only software, telecommunications, and payments and fintech companies. But those self-assessments often mask fragmented efforts that lack scale, focus or executive ownership.
A central problem, Uribe said, is leadership.
“Very few management teams at the CEO level have said this is a top three priority for the bank,” he said. “If you don’t have that top-down ambition, that top-down role modeling of behaviors, it ends up being very difficult for people in organizations to actually drive the impact that’s needed.”
The stakes are rising quickly. BCG projects that retail banks face a “toxic” mix of slower revenue growth and increasing costs. It forecasts revenue growth slowing to 2% to 4% annually from 2024 to 2029 as costs climb. In North America, pretax profitability already declined between 2021 and 2024 as operating expenses and loan-loss provisions rose.
Incremental cost-cutting will not be enough to reverse the profit squeeze. They also need to find new ways to raise revenue, which AI can help unearth. For example, they can utilize gen AI to improve customer engagement to cross-sell other products, according to Uribe. When a bank can see more deeply across a customer’s financial life, it can recommend ways to invest or cut costs – and earn commissions or referral fees.
So far, most banks have concentrated AI deployments in relatively safe use cases. Customer service is the most common entry point.
“One of the initial use cases that has been a perennial one that I’ve seen across the board has really been in the area of customer service,” Uribe said. “Using large language models to go through and mine the calls, literally the call transcripts, to understand what were the various things that were discussed, and classify that. There’s an element to really understand the drivers of call volume so you can go in and tackle them in the process.”
Banks are also experimenting with AI-assisted chatbots, virtual agents that help human representatives respond faster, and tools that search internal knowledge bases during live customer interactions. Fraud detection, know-your-customer (KYC) and anti-money-laundering (AML) processes are also seeing early adoption, including automated drafting of suspicious activity reports, Uribe said.
Other use cases are emerging in commercial credit, where AI can help draft credit memos, surface key risks and improve consistency, and in sales and wealth management, where advisors are being equipped with real-time insights about client portfolios.
Yet these applications remain narrow compared with what BCG calls an “AI-first” retail bank. In that model, AI is not just a support tool but a core operating logic that reshapes products, processes and customer relationships. AI agents, in particular, are expected to play a central role.
“AI agents are systems that autonomously observe, plan, and act toward goals,” the BCG report said. “Their integration into business workflows enables autonomous reasoning and action across the value chain – shifting AI from a passive advisor to an active operator.”
BCG estimates that AI agents already account for 17% of total AI value across industries this year and could reach 29% by 2028. For banks that move early, the payoff could be substantial. AI-first retail banks could achieve cost bases 30% to 40% lower than peers and increase profits by 30% or more, according to BCG’s analysis.
Surprise: Most banks not yet mobile-first
The hesitation is not only technical. Banking remains a heavily regulated industry, and executives worry about model risk, accountability and customer trust.
“There are regulators who are watchful and careful,” Uribe said. “It’s a matter of finding ways to implement it safely and securely and making sure that you don’t have hallucination issues.”
That, he said, requires robust AI governance, clear ownership, and decisions about where humans must remain in the loop. Many banks are still grappling with those questions while also finishing earlier digital transformations.
Uribe drew a parallel to mobile banking, which many institutions still struggle to execute well. BCG scores their efforts from zero to 100, where 100 means their mobile app gives them a competitive edge over rivals.
“You would think (most banks are mobile-first),” he said. Global banks score around 80 and top U.S. banks hover at 60. “But the average regional bank in the U.S. – we have a little over 4,000 – among the top, let’s say, five to 20 banks score at 30. That gives you a sense of how far behind they are.”
This finding is startling given that a recent BCG consumer survey revealed that the main channel customers interact with their bank is through mobile, with nearly seven out of 10 of them using this channel, Uribe said.
But AI can help them catch up. With the right leadership and ambition, lagging banks could leapfrog rivals rather than simply falling further behind.
“If you have that right, top-down ambition and that right, top-down leadership,” he said, “this probably does create an opportunity for some of the ‘laggard banks’ to leapfrog ahead of competitors and catch up in a way that maybe wouldn’t have been possible before.”
“This is an exciting time,” Uribe concluded. “The financial institutions that dream big and align resources and restack priorities are the ones that I think are going to create outsized advantage.”
