Inside the 2026 AI Index

Stop us if you’ve heard this before: JPMorganChase is the top bank on AI, this year’s Evident AI Index for Banks shows. It’s the fifth time America’s largest lender has bested its competition on AI prowess.

Capital One again grabbed second in our ranking of 50 of the world’s largest banks. RBC – the top non-U.S. lender in the Index – held its podium spot. And Australia’s CommBank stayed fourth.

Squint and you could mistake the leaderboard below for a rerun. But underneath what looks static is an industry moving faster on AI than ever. The rest of the world may be busy debating whether this tech – as some alarmist leaders have said – could someday kill us all. Banks, our Index shows, have a more immediate fear: that not moving fast enough on AI could kill their businesses first.

Their urgency shows up in the ranking. The average lender improved its Index score by 26% compared to last year, nearly three times the pace seen over the last three editions. The field moved so quickly that even banks that substantially improved slid down the table because rivals went even faster.

As every bank has started rolling out AI, simply having the technology no longer counts for much. “I’m going to do it, everyone else is going to do it,” JPMC CEO Jamie Dimon said earlier this year, arguing that AI use alone wouldn’t afford the bank much of an advantage. The premium now is doing the right things with it so that it can scale across a firm and through each line of business a bank has. On that, Dimon said, “you want a head start.”

What makes a leader in this Index, then, is proving that a bank can turn its head start into bottom-line impact. For that, the leaders have called in the cavalry. So-called AI enablement staff, the people charged with getting the technology working for the business, grew 21% this year (see: “New AI talent war,” The Brief, Sept. 3). AI product management staff grew 10%. And model risk talent – the people making sure more powerful tools can be trusted – grew nearly 10%, too. At the same time, lenders have put in place the AI platforms that let different parts of the business reuse models, code and guardrails so AI scales without its benefits eroding.

The results, at long last, have started to come through. Twelve banks in our Index report AI’s actual or targeted value, up from eight last year. CommBank and TD Bank set their first targets since our last ranking. Bank of America, meanwhile, reported it was in the black from its Gen AI investment this month (see: “Stat of the week,” The Brief, Oct. 1).

“Moving AI from pilots to large scale production is one of the defining leadership challenges facing organizations today,” wrote Saul Van Beurden, head of AI at Wells Fargo, a bank that jumped a spot to enter the top five. The rise, he said, was possible because of the people “building the AI platform and capabilities, running the AI governance and safety, and training and adoption needed to turn AI’s potential into real impact.”

So what actually separated the banks at the top – and those moving up – this year? The Index assesses banks on the AI talent they’ve amassed, the scale and impact of the use cases they’ve rolled out, their innovation efforts, the tech leadership of their top executives and the guardrails they’ve set up to govern this technology effectively.

Here’s a whiparound look at the biggest moves leaders made to climb the ranks.

🏅LEADERS OF THE PACK🏅

#1 JPMORGANCHASE: IN THE BLACK = GOLD

Talent: #2 | Innovation: #2 | Leadership: #1 | Transparency: #1

JPMC ended last year making back the $2 billion it spent on AI development. Since then, the Index leader has shown the blueprint for what comes next. The bank doubled down on putting its businesses in charge of AI, embedding data and AI chiefs into individual lines of business and letting them work directly with Jamie Dimon’s top lieutenants to rethink the way work gets done (see: “AI remakes the org chart,” The Brief, Feb. 19). Dimon says that’s already helped some parts of the operation cut headcount by up to 40% (see: “Between the bottom lines,” The Brief, July 16). But JPMC isn’t simply banking those savings: Overall headcount is up as it looks to do more. In private banking, for example, Derek Waldron, the bank’s chief analytics officer, says AI-led reorganization could help the bank cover 50% more clients.

#2 CAPITAL ONE: HOMEGROWN CHALLENGER

Talent: #1 | Innovation: #1 | Leadership: #40 | Transparency: #18

Capital One spent years building its tech stack to give the bank cleaner data, then recruiting the Index’s top bench of talent to put it to work. In 2026, the bank showed a blueprint for moving beyond the agentic pilot and industrializing agentic tools. The bank took MACAW, the multi-agent framework behind its Chat Concierge car-buying tool, and pushed it into other parts of the bank, including call centers, where it outperformed the Gen AI tools already working there (see: “Stat of the week,” The Brief, March 26). That kind of reuse works because the bank’s enterprise platform gives agents the information and guardrails so they act more like Capital One employees than standalone chatbots. “Agentic AI now must be treated as an end-to-end system,” said Rashmi Shetty, the bank’s VP of enterprise AI, last month. “Providing agent context becomes that much easier with a very strong data foundation.”

#3 RBC: OUT OF THE LAB

Talent: #8 | Innovation: #5 | Leadership: #2 | Transparency: #3

RBC spent a decade building Borealis into one of the sector’s deepest AI research shops. In 2026, the Canadian lender pulled that expertise closer to the business, creating a new AI group led by Bruce Ross and tasked with turning more of what Borealis invents into things RBC can actually use. ATOM, the bank’s foundation model, may be the clearest example. Announced publicly in 2025, the model – which is trained on “billions of client financial transactions” – has the institutional knowledge of the bank baked in and is now being used across 15 products and processes. That includes credit adjudication and customer personalization. The bank is betting that reuse can help it reach the roughly $700 million in AI value it’s promised to investors by next year. “We spent the past decade building the foundation,” CEO Dave McKay told investors this year. “Now we’re tapping into that leadership to deliver.”

📈ON THE RISE📈

• #8 Citi (+4) rode new leadership and a new platform to the top 10, climbing four spots in this year’s Index. In 2026, David Griffiths took the helm as the bank’s group head of AI and rolled out Arc, a platform that streamlines how the bank deploys agentic tools across its lines of business. “We’ve built industrialized infrastructure to make AI agents a core part of how we serve our clients,” Griffiths said. “We’ve got agents talking to other agents as part of doing business.”
  
  
• #10 TD Bank (+3) rolled out its first agentic tool – which saves 15 hours per mortgage – en route to a top-10 finish (see: “House rules,” The Brief, May 28). That’s contributing to the $137 million the bank has made already this year from AI (see: “In the news,” The Brief, Sept. 3). The bank is also getting serious uplift from developer productivity tools, where it has a 92% uptake and is improving its capacity by up to 40%. “The single most impactful AI program that we are running right now is our software development program,” said Leo Salom, head of U.S. Banking at TD, last month.
  
  
• #36 Westpac (+9) climbed out of the basement after raiding rival CommBank for AI talent and building the foundations for its own AI push. The bank put CommBank alumni Andrew McMullan, Dan Jermyn and Maggie Shi in senior AI roles. The bank is now rolling out Adapt, a new enterprise data platform that pulls 285 systems into one environment that will sit under the bank’s AI buildout (see: “Adapt or die,” The Brief, Sept. 17). “Every digital experience, every insight and every AI capability depends on the quality of this foundation,” McMullan said.

If there were a “word of the year” in banking in 2026, it would be reimagination.

Lloyds (#15) is “reimagining how we operate by harnessing the full potential of AI.” CommBank (#4) is using “AI to reimagine banking for customers.” UBS (#6) is using the tech to “reimagine the way we work.” Who says bankers aren’t creative?

But there’s a reason for its use (or overuse). We analyzed more than 1,100 use cases rolled out by banks and graded them on their impact – how much they were returning to the bank and how widely they were being or could be used. The tools that got the most points – including the three below – weren’t just doing the same job faster. They showed how the job itself could be done differently with AI.

JPMorganChase’s SpectrumIQ

• What it does: The tool pulls together research, market data and risk information to help investors decide what to buy or sell across the firm’s institutional equities and FX trading businesses. Then it helps execute the trades themselves.
  
  
• Impact: Nearly 75% of equity trading and 85% of FX trading in its asset management arm is automated. Equity automation has saved clients $4 billion in trading costs since inception.
  
  
• What got reimagined: All of JPMC’s (#1) investment workflows. Powering Spectrum IQ is data on 90,000 securities and 22 million documents, a figure that grows by 7,000 new broker research reports per day. The infrastructure, then, can be used to power other tools, like those in research, portfolio construction or proxy voting.

UBS’ STAAT Insights

• What it does: The tool scans thousands of internal and external signals for information that a wealth advisor should know about a client and sends it to them before a meeting. It also looks on its own for opportunities to engage with clients.
  
  
• Impact: Nearly 90% of advisor teams are using it, saving some 1,200 hours of meeting preparation each week.
  
  
• What got reimagined: How UBS (#6) turns data into action. By bringing internal client information and outside signals together, the bank can have AI continuously look for things worth acting on rather than having advisors search for them. Unified data like that now underpins UBS’ AI platform Claves, which underpins tools being built across the business.

BMO’s Personalized Client Offers

• What it does: The tool uses customer data to work out which product or offer will be most relevant to someone and then serves them that recommendation.
  
  
• Impact: The approach generated approximately $60 million in annualized revenue in 2025.
  
  
• What got reimagined: How BMO (#20) sells to customers. Instead of running separate campaigns for different products, the bank built a system that continuously works out what customers might need or want next. This “Next Best Offer” feature drove a 3% lift in sales revenue, the bank said. And the data can, in turn, inform other experiences customers have with the bank across its channels.

For a while now, CEOs have had to sound fluent on AI to investors. As the tech has scaled, they now need to be able to speak the language of AI back to the people in their organizations actually building it.

“If your CEO is not the one signing off on your fundamental transformation…you’re missing an opportunity,” said Charles Holive, chief AI officer for BNP Paribas’ (#16) corporate and institutional banking arm. “If I have a leader focused just on efficiency plays, they will miss revenue growth opportunities,” Kristin Milchanowski, chief AI and quantum officer at BMO (#20), said last month.

For two years running, five CEOs (seen below) have scored top marks for what they said to investors and the media and how closely they tie AI to the bank’s strategy. These are some of the most interesting things they’ve had to say this year.

“Reasoning, the access to tools, the amount of context that you can put into it - your token costs do not scale on a linear basis.”

– Matt Comyn, CEO at CommBank, preempting the AI cost reckoning that came to banks this summer, June 2026 

“[AI agents] may lead to us winning and losing in big areas.”

– Jamie Dimon, CEO at JPMorganChase, recognizing that agents could redraw the competitive map as customers trust them more, January 2026 

“One Transformation is not a cost-cutting program. It is a redesign of our operating model, the consequence is that cost comes down.”

– Héctor Grisi, CEO at Santander, explaining the overhaul that led to the bank reporting the quarterly value of AI’s financial impact, February 2026

“AI gives us places to go we haven’t gone.”

– Brian Moynihan, CEO at Bank of America, making the case for judging AI by what it lets a bank do rather than how much it can cut, April 2026

“We look at [agentic AI] from three lenses – personal agents, team agents and enterprise agents (which support complex agentic workflows).” 

– Tan Su Shan, CEO at DBS, outlining how the bank plans to turn individual AI use into enterprise change, May 2026

We can’t predict every turn AI will take in 2027. But as this year’s Index dust settles, we’re already seeing banks make big bets on where the next advantage may come from.

Over the next few weeks, we’ll dig into several and highlight the signals worth watching. First up: The chip race comes to banking.

The bet, explained: To scale AI, banks don’t just need powerful AI models; they need enough computing power to run them. Today, they either rent that power from cloud or specialist providers or buy their own chips – like GPUs, the kind of chip Nvidia makes. That has worked while AI use is relatively contained and led by humans. But as always-on agents become a bigger part of life, demand for the already-expensive chips could go through the roof. Banks are laying the groundwork now – lining up computing power and seeking out chips that can do more with it – so their ability to scale doesn’t get capped by hardware.

What we’re seeing: Trading firms show how quickly AI’s appetite for computing power can grow. Trading firm Jane Street, for example, is snapping up GPUs to support its AI buildout. It already has tens of thousands of GPUs on hand to train and run its models and plans to grow that to the hundreds of thousands, Ron Minsky, the firm’s co-head of tech, said. Banks are also planning for that rising demand. John Sharratt, global head of technology and infrastructure at Standard Chartered, said it had secured “hundreds” of GPUs for projected AI demand through 2027. JPMorganChase, meanwhile, is starting to spread its bets: This summer the bank signed a deal with SambaNova, a startup that makes a different kind of AI chip than Nvidia, to run some of its AI models.

What’s next: Banks’ first move will be to try to squeeze more out of the chips they already have. China Merchants Bank showed as much by cutting inference costs by more than 60% for comparable model usage by reconfiguring how existing hardware is put to work. But even that efficiency will only go so far. As AI grows, lenders will need to decide how much more computing power to buy and how far ahead to plan. Getting that calculation right could make a bank jump in next year’s Index. “One of my greatest failures has been… predicting how many GPUs we would need,” said Iain Dunning, head of AI at Hudson River Trading. “You’re constantly playing catch up.”