It's not the AI, stupid

The recent string of layoffs across the banking industry is being blamed on AI. There’s a different acronym that should come to mind: BS.

Look at all the cuts piling up, in many cases with AI on the wanted poster. Nordea plans to cut 5% as it expects AI to deliver cost savings. HSBC says up to 10% of its staff could be on the chopping block in the next several years. 

Tech is being framed. Faced with investor pressure over a fragile market, banks have learned Silicon Valley’s old trick: Put AI in the headline and people stop asking hard questions about costs, growth and management. 

The fact is banks aren’t nearly good enough at AI yet to explain cuts this deep. For now, AI is helping lenders go faster. Tools are helping squeeze more output from the same people – 100,000 hours for developers at Citi, 1.2 million hours firmwide at CIBC. But they create a productivity paradox: All that extra time still can’t replace the roles altogether. Even at JPMorganChase, where 150,000 active LLM Suite users save four hours each week, CEO Jamie Dimon says it’s “not something you can directly translate into headcount reduction” (see: “Show me the money,” The Brief, Feb. 26). 

When firms have tried to replace workers so far, they’ve largely failed. Klarna had to hire back customer service staff after its cost-cutting AI push led to “lower quality” interactions. On a smaller scale, CommBank walked back job cuts brought on by a customer service bot after complaints climbed. In each case, technology could handle pieces of the job, but it couldn’t make the cost savings stick or contain the new risks it introduced. People, they found out, remained essential.

What could justify such large-scale cuts is what banks are only just beginning to do: Rebuilding the work itself around AI. That means not just layering AI on top of what a bank already does, but “re-engineering processes for human and AI collaboration,” as DBS CEO Tan Su Shan described it this month (see: “DBS’ new success metric,” The Brief, March 12). 

They’re nowhere near that yet, at least not at scale. This past week, Goldman Sachs CEO David Solomon laid out six “initial workstreams” where it wants to rethink around AI, ranging from KYC and client onboarding to vendor management (more on that strategy in “Use Case Corner” below). After stripping those processes back to the studs, it’ll have to repeat the same process for every other area of the bank.

Until that happens, banks aren’t really cutting jobs because of AI. They’ve just found a perfect scapegoat.

The conflict in the Middle East will be “short-lived, meaning weeks, not months,” BBVA CFO Luisa Gómez Bravo told investors last week. That may or may not be true, but there’s one thing you can say for sure: The war’s impact on the Gulf’s AI buildout in banking could last years.

For banks – both in the Gulf and globally – trying to turn the area into an AI Hub, the conflict poses a long-term threat. The AI talent needed to make that strategy work may no longer want to come or stay.

For years, Gulf states have laid the groundwork for an AI boom. Massive sovereign investments to lock in cheap energy, aggressive data center buildouts and state-backed strategies to court foreign firms have been used to position the region as a base for finance and technology. For banks, places like Dubai and Doha were pitched as new Genevas, stable bases for capital that could support AI growth.

Those efforts were working. AI use in banking could lift the GCC’s GDP 14% by 2030, research firm Finastra projected last year. Global banks came knocking: Standard Chartered gets about 8% of its revenue from the region now, JPMorganChase analysts said this month. HSBC, meanwhile, opened a new wealth center in the UAE last September to take advantage of an area “attracting more net inflows of millionaires than any other country in the world,” Mohamed Al Marzooqi, the bank’s UAE chief executive said.

In this “If you build it, they will come” approach, talent was the missing piece. As of Q4 2025, no Middle Eastern cities cracked the top-10 global hubs for AI talent across the 50 banks tracked by Evident. Last year though, both the UAE and Saudi Arabia cracked the top-20 countries for AI talent, according to a report from the International Finance Forum.

War may well reverse that progress and force the AI talent drawn to the region to second-guess that choice long-term. Citi and Standard Chartered shut some UAE offices this month and evacuated employees. HSBC closed its branches in Qatar. Even when the conflict ends, hedge fund Millennium is bracing to handle employees that don’t want to go back to Dubai by relocating them to Jersey. Even if temporary, those moves make it harder to recruit and retain the engineers banks need on the ground – especially since data sovereignty rules in the region limit remote work possibilities.

That’s not to say the Gulf’s AI push is undone. The capital is still there and the infrastructure largely remains. When the conflict ends, banks will face a different kind of rebuild: Convincing AI talent the region is still a place worth building in.

How much less frequently humans had to edit call notes when Capital One swapped a single LLM for a multi-agent system in its call center summarization tool, Milind Naphade, the bank’s SVP of AI foundations and Ritesh Soni, managing VP of the retail bank, told the crowd at last week’s Nvidia GTC conference. Instead of using one model to do everything, the bank split the job across specialized agents: An “understanding agent” works with a “reasoning agent” to list the facts. An “evaluator agent” then checks that passage for accuracy. And finally an “explainer agent” puts the summary in a format that’s easy to read. AI-written summaries were rated better on accuracy, completeness and readability than human work by annotators at the bank.

Bigger picture: Capital One is starting to treat some early AI tools as legacy tech as it makes more agentic bets. The bank recently retired a developer AI tool, Catherine McGarvey, SVP of developer experience, said in an interview last week. “We have this mindset of, it doesn’t matter how far we’ve come,” she said. “If it’s the wrong call or we learn something new that changes our behavior, we can [change].”

David Rice was named HSBC’s first chief AI officer. Rice has been with the bank for almost two decades and was previously COO for the corporate and institutional banking unit. The move “provides clear enterprise leadership for AI adoption,” the bank’s release said. It reflects a growing push among banks to adopt a hub-and-spoke structure that lets AI strategy be set at the group level and leaves execution to the lines of business. “David will be instrumental” in empowering the firm’s employees to use AI to create personalized experiences for customers, CEO Georges Elhedery said. An HSBC spokesperson confirmed that Rice will report to both COO Suzy White and CIO Stuart Riley.

Shane Artis is now head of AI transformation and sustainable technology across non-financial risk technology at Morgan Stanley. His remit includes looking after business-facing agentic AI development, he wrote on LinkedIn.

JPMorganChase is growing its LLM Suite team, hiring senior executive directors and VPs to work on the firm’s AI platform. 

Scotiabank is hiring a head of AI and agentic engineering who will lead “the design, delivery, and enterprise-wide adoption of the full spectrum of applied AI capabilities,” across the bank. 

Google DeepMind hired Bridgewater Associates’ chief scientist and head of AI, Jasjeet Sekhon, to be its new chief strategy officer. Sekhon helped build AIA labs, which is the hedge fund’s AI research and investment lab.

In its annual report, Goldman Sachs laid out six areas for full AI-led transformation: Vendor management, sales enablement, KYC / customer onboarding, lending, enterprise risk and regulatory risk. In this week’s “Corner,” we look at a best-in-class use case in three of those areas which preview what true transformation can look like in those domains:

SALES ENABLEMENT: The clearest avenue to a bottom-line impact.

• Use case: Coach AI 
• Bank: JPMorganChase
• Why it’s ripe for transformation: JPMC’s 10,000 financial advisors used the tool – which offers real-time tips on how to guide clients – and closed 30% more deals as a result, the bank said. That came with the tool still acting like a helper: “It is a bit annoying…asking the agent to continuously complete tasks,” said Ajay Mehta, who leads the bank’s Connect Platform. Translation: Humans still drive the process. Once a bank’s systems go from fetching information faster to doing prep work and follow-ups on its own, 30% growth will be the floor.

KYC / ONBOARDING: Any extra time spent on the process of verifying documents takes money out of the bank’s pocket.

• Use case: Source of Wealth Assistant
• Bank: Bank of Singapore
• Why it’s ripe for transformation: The OCBC-owned private bank cut the time it took employees to draft “source of wealth” memos – which explain how prospective clients made their money – from 10 days to one hour. That’s just one small part of the KYC workflow. Leading banks can automate about 40% of the whole process (see: “KYC on auto,” The Brief, Dec. 11). But if a bank can redesign that process wholly around AI, they will move from doing KYC faster to making big chunks of the process run on its own.

VENDOR MANAGEMENT: An unsexy cost center at banks that eats up billable hours.

• Use case: Contract review assistant
• Bank: BNY
• Why it’s ripe for transformation: BNY’s tool, developed with OpenAI, cut the length of contract review cycles by 75% for more than 3,000 contracts. That solves one bottleneck. But true AI transformation in vendor management will come from agents managing the outside vendors more holistically, taking data on risks, spending, monitoring and more that exists in disparate parts of the business and bringing it together into one streamlined process that gives the bank more control over the economics and the risks that come with bringing on new partners.

“There is no consensus on what AI will mean for the labor market – particularly for entry-level white-collar roles. The truth is, no one knows with certainty. In the near term, there are roles we know are in clear demand, and pay well: skilled trades, especially the ones building the physical infrastructure of AI, like data centers, power systems, and electrical grids.”

– Larry Fink, CEO at BlackRock, in his shareholder letter, March 23

Starling Bank this week launched the “U.K.’s first agentic AI financial assistant,” a tool it says can help app users manage their money and get personalized financial advice. The neobank’s customers can speak or write to the assistant, which can take actions – like automatically transferring money between accounts each month based on a savings goal a customer sets out. Personal finance has become a new battleground, but not all AI tools here are the same. NatWest, for example, launched Cora Spending Chat, a customer-facing agentic tool in its app which lets customers interrogate their spending. American Express (#4 in the Evident AI Index | Payments) is launching an agent that “provides deep spending insights,” CEO Stephen Squeri wrote this week in his shareholder letter. Lloyds signaled in November that it would also launch an agentic financial assistant this year; the question is whether it’ll simply surface insights in a new way or actually take actions on its customers’ behalf.

Palantir won a three-month trial contract from the Financial Conduct Authority, which will see the U.K.’s financial watchdog using the U.S. tech company’s software to help tackle financial crime. Palantir has seen other wins in the financial sector in recent months, with the Peter Thiel-backed firm partnering with Swiss Re in December to help the reinsurance company build an AI platform. 

Bank of America rolled out a new AI tool for investment advisors to help prepare, summarize and take further action around client meetings. “Meeting Journey” was rolled out to 25,000 advisors and is saving them up to four hours per meeting, the bank reported. It plugs into Salesforce to prepare meeting notes and Zoom to create summaries. It’s a late arrival: JPMorganChase’s Advisor Copilot and Morgan Stanley’s Debrief tools have similar functionality and were rolled out roughly two years ago.