When Standard Chartered’s CEO Bill Winters stood before investors in Hong Kong recently, he described thousands of his own employees as ‘lower-value human capital’ that would be replaced by artificial intelligence.
The backlash was swift. Regulators called. The staff were furious. Winters apologised on LinkedIn within days, insisting he had only meant to flag which types of tasks were at risk from automation, not pass judgement on the people doing them.
The episode was awkward and poorly handled. But it also cracked open a conversation that the entire global banking industry had been conducting quietly behind closed doors. AI is coming for a significant portion of the financial workforce, and no one has quite agreed on how to talk about it, let alone manage it.
A reckoning long in the making
Banks have always run on paperwork. Behind every mortgage approval, every wire transfer, every new client account is a long chain of human checks. Someone verifying an identity, another person reconciling a transaction, a third ensuring that everything meets the regulatory requirements of whichever country the money happens to be moving through. For decades, this work was simply too complex and too context-dependent to be handled by machines.
That is no longer the case. The new generation of artificial intelligence (AI), particularly the large language models and agentic systems that have emerged in recent years, can read documents, cross-reference data, flag inconsistencies, and generate compliance reports at a speed and scale no human team can match. For banks, which operate some of the largest back-office workforces in the world, this is both an extraordinary opportunity and a serious problem.
Research by Citigroup found that around 54% of all banking roles have a high potential for complete automation. A further 12% could be significantly augmented by AI tools. McKinsey put the financial upside of this shift at a 25% reduction in operating costs, with roughly 30% of all work hours in finance and insurance fully automatable by 2030.
Citigroup’s own models suggest the industry could add as much as 170 billion dollars to its collective profits by 2028 simply by deploying these systems at scale. These are extraordinary numbers. They are also, for hundreds of thousands of people currently employed in those roles, a source of genuine anxiety.
What the banks are doing
The scale of planned workforce reductions across major institutions is significant, though the language used to describe them varies widely.
Standard Chartered has announced it will cut between 7,800 and 8,000 corporate function roles by 2030, roughly 15% of its support staff. The cuts are concentrated in operational hubs in Chennai, Bengaluru, Kuala Lumpur, and Warsaw, where large teams handle back-office and middle-office functions.
Regulators in Hong Kong and Singapore contacted the bank for clarification, with Hong Kong’s monetary authority specifically asking whether AI was being used as cover for straightforward cost-cutting.
HSBC, which employs more than 200,000 people worldwide, is reportedly weighing a restructuring that could affect as many as 20,000 roles over the next three to five years. Chief Executive Georges Elhedery has been more careful with his framing than Winters, urging staff to stop resisting the technological shift and to think of it as a transition rather than a termination.
The bank has appointed its first dedicated chief AI officer and is pushing generative AI tools into customer onboarding, risk monitoring, and wealth management. The overall message is collaborative, though the underlying numbers tell a harder story.
JPMorgan Chase, the largest bank in the United States, has taken a different approach to communicating the same reality. Chief executive Jamie Dimon has said plainly that the bank will hire more AI engineers and data scientists while reducing its intake of traditional banking staff.
Dimon described Winters’ controversial remarks as merely ‘inartful’ rather than wrong. JPMorgan manages its headcount reduction largely through natural attrition, relying on the fact that roughly 25,000 to 30,000 employees leave the firm voluntarily each year. This gives the bank room to reshape its workforce gradually without mass redundancies.
Citigroup has set a target of cutting around 20,000 roles by 2026, focusing on middle-office and operational support functions. At the same time, the bank has given AI tools to 40,000 software developers and rolled out internal AI platforms to nearly 180,000 employees across 83 countries.
Mandatory training in prompt-writing, the skill of giving AI systems clear and useful instructions, has been extended to 175,000 staff. These tools have already freed up around 100,000 hours of weekly capacity within the bank’s technology teams.
Goldman Sachs has framed its strategy around what president John Waldron calls the ‘human assembly line’ problem. Banks, he argues, have long worked like factories, with people passing information down the chain from one specialist to the next. AI agents, which can work continuously without a break and handle multiple tasks simultaneously, are now capable of performing much of that assembly-line work autonomously.
Goldman is deploying AI agents built on Anthropic’s Claude model in two areas: trade and transaction accounting, where the agents reconcile millions of financial records that previously required large teams of accountants, and client due diligence, where they process the extensive documentation required to verify new institutional clients under anti-money laundering and know-your-customer rules.
The ‘AI-Washing’ question
Not everyone accepts that the scale of these changes is driven purely by technology. Critics, including several prominent economists and venture capitalists, have noted that many large companies became heavily overstaffed during the post-pandemic hiring surge, and are now using AI as a convenient justification for reductions they would have made regardless.
Thomas Malone, a professor at MIT’s Sloan School of Management, has observed that for companies facing sluggish revenue or strategic missteps, blaming job cuts on automation ‘definitely makes for a better story’ when addressing investors than admitting to poor planning.
Marc Andreessen, the technology investor, made a similar point, suggesting that generative AI has become a ‘silver-bullet excuse’ for shedding excess headcount while keeping shareholder confidence intact.
The banking sector, however, has a stronger structural case for automation than most. Unlike technology firms, which can scale their products to millions of users with little additional labour, banks are required by law to check every transaction, verify every client, and document every decision.
This compliance burden has historically meant that revenue growth required proportional headcount growth. AI systems that can shoulder much of this burden without additional staff represent a genuinely transformative shift in the economics of the business, not just a story for investors.
Fresh out of college, all out of jobs
One consequence of this shift that has received less attention than headline redundancy numbers is what is happening to entry-level hiring. Data from municipal financial reports in major cities points to the emergence of what analysts are calling a ‘low-hire, low-fire’ economy. Overall layoffs have remained relatively modest, but firms are simply not replacing the junior positions that AI tools have begun to absorb.
The effects are already visible in graduate employment data. For the first time on record, unemployment among recent university graduates has in some months exceeded the unemployment rate for young adults without a degree.
Tasks that were once carried out by trained junior bankers, including first-draft financial modelling, initial document review, and basic compliance checking, are increasingly handled by AI tools before a graduate would ever have had the chance to do them.
The National Foundation for Educational Research has warned that AI and automation could eliminate up to three million low-skilled jobs in the United Kingdom alone by 2035. The banking sector’s shift is one of the more visible early examples of a broader structural change in how white-collar work is organised, and who gets to do it.
Reskilling models worth studying
Amid the disruption, some institutions have moved beyond announcements and built programmes that appear to be working in practice.
DBS Group in Singapore has become something of a benchmark for how a bank can manage this transition without mass redundancies. Rather than announcing large-scale cuts and then scrambling to manage the fallout, the bank began by freezing external hiring for roles it identified as vulnerable to automation.
At the same time, it launched a comprehensive retraining programme for around 13,000 employees, with more than 10,000 already having completed their initial AI and data skills coursework.
The approach to individual roles has been thoughtful. Bank tellers, for instance, have not been let go. Instead, they have been retrained to manage interactive video teller machines and transition into digital relationship management, where human connection and judgement remain valuable.
In Singapore’s customer contact centres, DBS reskilled roughly 500 staff members into 13 new job categories, ranging from digital content creation to customer experience design. Over time, the number of traditional call centre agents fell from 300 to 55, but through reassignment rather than termination.
DBS also invested in AI-powered internal tools to support this process. OneBot handles HR and IT queries around the clock, reducing the administrative burden on staff. JIM, the Job Intelligence Maestro, screens candidates and predicts which employees are at risk of leaving, allowing managers to intervene early.
IGrow, a personalised career development platform, identifies skill gaps for individual employees and recommends specific training pathways tailored to each person’s profile. This work sits within a broader national initiative coordinated by the Monetary Authority of Singapore to retrain 35,000 financial sector workers across the country’s three main domestic banks.
In the United Kingdom, Lloyds Banking Group has built what it calls an AI Academy to upskill its entire workforce of between 60,000 and 67,000 people. Chief Executive Charlie Nunn has articulated the programme’s philosophy clearly. AI replaces tasks, not roles, and employees who learn to work alongside AI will eventually replace those who do not.
The financial case for this investment has been documented with unusual transparency. In 2025, early generative AI deployments at Lloyds, including a tool that reduced the time needed to categorise customer complaints from five minutes to one second, delivered a direct saving of 50 million pounds. That figure is projected to exceed 100 million pounds in 2026 as the bank expands its use of autonomous AI agents across more processes.
Lloyds has tied these efficiency gains to a specific curriculum built around skills that remain difficult for machines to replicate: critical thinking, ethical reasoning, the ability to question and verify AI-generated outputs, and genuine human empathy in client-facing situations. Training completion figures are published quarterly, and the bank maintains active dialogue with its main recognised trade union, Accord.
The relationship has not been without friction. In early 2026, Lloyds faced significant criticism after it emerged that the bank had used aggregated salary and spending data from 30,000 staff bank accounts during pay negotiations, arguing the data showed its lowest-paid employees were in a better financial position than the general public. Nunn apologised in an internal town hall and ordered a review. The incident underscored that the ethical use of data, internally as much as externally, is a live concern as banks integrate AI more deeply into their operations.
Four things banks must get right
The early evidence from institutions that have handled this transition well points to several practical lessons.
Training needs to be mandatory and consequential, not optional and token. DBS, Citigroup, and Lloyds succeeded because upskilling was treated as a core business requirement rather than a voluntary development exercise. At JPMorgan, AI training is linked to career advancement. The same principle applies elsewhere: if employees have no incentive to complete training, most will not.
Experimentation needs guardrails. JPMorgan’s approach of giving staff access to AI tools within secure, pre-approved environments allows people to learn by doing without creating compliance risks. Banks that simply hand out access to external AI products without governance frameworks are inviting expensive errors.
Natural attrition is more humane and more sustainable than mass redundancies. JPMorgan and DBS have both demonstrated that a patient, attrition-led approach to headcount management avoids the labour friction, regulatory scrutiny, and reputational damage that blunter approaches invite.
Human oversight must be built into AI workflows from the beginning. The banking sector’s reliance on autonomous agents raises a genuine accountability problem. When a machine generates compliance outputs faster than any human team can audit them, but human professionals remain legally responsible for signing off on the work, the system is structurally vulnerable. Designing clear human checkpoints into automated workflows is not a concession to inefficiency; it is a basic condition of responsible deployment.
The global banking industry is now facing a choice between technological progress and workforce responsibility. The banks managing this transition most successfully have found that these goals are compatible, provided the commitment to both is genuine. The challenge now is for the rest of the sector to follow.