Morgan Stanley predicts a 10% workforce reduction as banks chase American-level profitability through automation. But even JPMorgan is warning about what happens when nobody understands how the systems work anymore.
By Shafaq | Dec 31, 2025
If you work in back-office operations, risk management, or compliance at a European bank, Morgan Stanley has some uncomfortable news: Your job might not exist in 2030.
According to a new forecast from the investment bank, European banking institutions could cut approximately 10% of their workforce — around 212,000 jobs — by the end of the decade as AI and digitalization sweep through the industry. The cuts will hit hardest in precisely the departments that banks have historically relied on to keep operations running: the unglamorous, labor-intensive work of processing transactions, managing risk, and ensuring regulatory compliance.
The driving force isn’t just cost-cutting for its own sake. European banks are trying to close a persistent profitability gap with their American counterparts, and they see AI as the path forward. Morgan Stanley analysts note that many banks have already achieved a 30% increase in efficiency through AI adoption and further digitalization. The next logical step? Reduce headcount to match.
The cuts are already starting
This isn’t a distant hypothetical. Dutch bank ABN Amro is planning to cut a fifth of its staff by 2028. At UBS, human analysts are being replaced by digital avatars that can handle routine client interactions and generate reports. Other European banks are reportedly exploring similar measures, though most remain cautious about announcing specific job cut targets publicly.
The focus on back-office, risk management, and compliance roles makes strategic sense from a bank’s perspective. These are areas where AI excels: processing structured data, identifying patterns in transaction flows, flagging potential compliance issues, and automating repetitive tasks that previously required armies of analysts.
Unlike customer-facing roles where human relationships still carry weight, much of the work in these departments involves exactly the kind of rules-based decision-making and data processing that large language models and machine learning systems can handle at scale — and often more consistently than humans.
Chasing American profitability
The push to cut costs reflects a broader competitive dynamic. European banks have long trailed their American peers in profitability, hampered by fragmented markets, stricter regulations, and higher operating costs. The cost-to-income ratio — a key efficiency metric — is traditionally higher in Europe, particularly in France and Germany, which Morgan Stanley identifies as the markets likely to see the largest wave of restructuring.
American banks, meanwhile, have spent years investing in technology infrastructure and consolidating operations. The result is leaner organizations that generate higher returns on equity. European banks are now trying to close that gap, and AI offers a plausible path to doing it without the painful political battles that would accompany cross-border mergers or aggressive branch closures.
The 30% efficiency gain that Morgan Stanley cites is substantial. In an industry where margins are often measured in basis points, cutting operational costs by nearly a third while maintaining or improving service quality would be transformative. The question is whether those efficiency gains actually materialize in practice, or whether banks discover hidden costs in implementation, change management, and the loss of institutional knowledge.
The warning from the top
Interestingly, even as banks move aggressively toward automation, senior leadership at some institutions is pumping the brakes on the hype. Conor Hiller from JPMorgan Chase — an American bank that’s been a leader in AI adoption — offered a notably cautious take in recent comments.
“The only thing we need to be very careful about is that people don’t lose their understanding of the basics and fundamental principles,” Hiller said. “Otherwise, we are accumulating a big problem for the future.”
That’s a remarkable statement from someone at a bank actively deploying AI across its operations. And it points to a real risk that often gets overlooked in discussions about automation: institutional knowledge erosion.
When you eliminate the analysts who understand how risk models actually work, who built the compliance frameworks, who know why certain processes exist in the first place — you create dependency on systems that nobody fully understands. And when those systems produce unexpected results, or when market conditions change in ways the models weren’t trained to handle, you’re left with a workforce that can’t troubleshoot because they never learned the fundamentals.
The 2008 problem, AI edition?
Hiller’s concern echoes warnings that emerged after the 2008 financial crisis, when it became clear that many banks had created complex financial instruments that even their own traders and risk managers didn’t fully understand. The models said everything was fine — until suddenly they weren’t, and nobody could explain why or fix the problem fast enough to prevent catastrophe.
AI-driven banking operations present a similar risk. If you automate compliance checks using machine learning models that flag suspicious transactions, but nobody on staff understands the underlying regulatory logic or can explain why a transaction was flagged — you’ve created a potential compliance nightmare. If your risk management is handled by AI that predicts defaults and sets credit limits, but your team can’t interpret the model’s reasoning or adjust it when market dynamics shift — you’re flying blind.
The difference is that this time, the complexity is baked into opaque neural networks rather than structured financial products. And unlike the pre-2008 era, when at least some people understood the math behind CDOs and credit default swaps, modern AI systems can be genuinely inscrutable even to their creators.
The political and social dimension
There’s also a political calculation that European banks will need to navigate. Cutting 212,000 jobs across the industry is the kind of move that attracts regulatory scrutiny and public backlash, particularly in countries like France and Germany where labor protections are strong and banking employment has traditionally been seen as stable, middle-class work.
Banks will need to manage the optics carefully — emphasizing retraining programs, natural attrition, and the creation of new roles even as they eliminate existing ones. They’ll argue that the alternative is worse: falling further behind American competitors, losing market share, and ultimately facing more severe cuts or even failures.
But the social contract around work is being rewritten, and banking is just one industry grappling with it. The promise of AI has always been that it would eliminate tedious, repetitive work and free humans to focus on higher-value tasks. The reality for 212,000 European banking workers will be more complicated.
What comes next
Morgan Stanley’s forecast extends to 2030, which in AI timescales is an eternity. The technology will continue evolving, regulatory frameworks will adapt, and banks will learn — sometimes painfully — which automation strategies actually work and which create new problems.
What’s clear is that European banking is heading for significant restructuring. The question isn’t whether AI will reshape the workforce, but how quickly it happens, whether the efficiency gains materialize as predicted, and whether banks can avoid Hiller’s warning about accumulating “a big problem for the future.”
For the 212,000 workers potentially at risk, the advice is familiar but cold comfort: upskill, adapt, and hope your particular role proves harder to automate than the forecasts suggest.
For the banks themselves, the path forward involves a delicate balance: cut costs aggressively enough to compete with American peers, but not so aggressively that you hollow out the expertise needed to understand what your AI systems are actually doing.
It’s a bet that efficiency gains will outweigh the risks of institutional knowledge loss. By 2030, we’ll know if they were right.
