About 1,000 of those roles are already gone. The rest are due out by the end of the year, and roughly 60% sit in support functions rather than in branches or on trading desks. Barclays employs around 100,000 people, so the cut is close to 5% of the workforce.
What the other European banks have announced
Barclays is the loudest case, not the only one. European Business Magazine reported over the weekend on a set of plans that add up to the largest white-collar reduction in the sector since the financial crisis.
HSBC has committed to roughly 20,000 job cuts alongside its deepest removal of senior executives in more than fifteen years. ABN Amro wants full-time staff down by 20% by 2028 through automation. Standard Chartered plans to remove close to 8,000 back-office roles by 2030, around 15% of central functions. UniCredit intends to bring its back-office share from 50% of total staff down to 25%.
The roles being removed cluster in the same places across all four banks: transaction monitoring, compliance checks, regulatory reporting, risk management and routine software development. These are tasks that large language models handle well and cheaply, because the work is text in, structured text out, with a human reviewing the exceptions.
The forecast that doubled in a year
Morgan Stanley has raised its projection for European banking employment. Its earlier estimate was that AI would remove 10% of sector jobs by 2030. The bank now puts the figure at up to 20%. That second number is roughly 400,000 positions, which implies a sector workforce near two million people.
A forecast is not a redundancy notice. What makes this one worth reading is the timing: it moved from 10% to 20% in about twelve months, while the banks themselves were announcing concrete cuts rather than pilot programmes.
The age structure is doing quiet work
European banking is older than most industries. About 27% of staff across the sector are over 50. At Italy's Banco BPM the share of employees over 50 reaches 60%, and at French lenders roughly 20% are over 55. Only about 15% of the sector's employees are under 30.
That distribution gives management an option it would not have in a younger industry. Natural attrition and early retirement can absorb a large part of a 20% reduction without formal dismissals, which keeps restructuring costs and public conflict lower. It also means the people leaving first are the ones with the most institutional knowledge, and the people staying are the ones who will be asked to work alongside the systems that replaced their colleagues.
The number that does not fit
An ECB survey of around 5,000 eurozone companies found that only 15% named cost cutting as their main reason for adopting AI. The rest pointed to other motives, from product quality to capacity. The same survey found no sign of AI-driven net unemployment across the wider economy.
Both things can be true at once. Banking is a sector where margins are thin, headcount is the largest cost line, and the work is unusually well suited to language models. A general survey of 5,000 firms across manufacturing, retail and services will not pick up a cull concentrated in the back offices of a dozen large lenders.
What EU law covers, and what it does not
The EU AI Act restricts several things a bank might want to do, but it does not stop a bank from reducing staff. Emotion recognition in the workplace has been banned since February 2025 under Article 5. The transparency rules of Article 50, which require disclosure when people interact with an AI system, have applied since August 2026. Providers of general-purpose models have carried documentation and risk obligations since August 2025.
AI systems used in hiring, evaluation, task allocation or termination fall into the high-risk category. Their obligations were pushed to 2027 and 2028 by the AI Omnibus regulation that entered into force in July 2026, with lighter administrative rules for smaller providers.
None of that creates a duty to justify a redundancy to a regulator. Collective consultation, works councils and national labour law still govern how the reduction is carried out, country by country. GDPR gives an employee the right to ask what data an employer holds and to object to decisions taken solely by automated means, which matters more as more of the assessment work moves to software.
Two timelines running at once
Barclays is operating on a twelve-month timeline: 1,000 already out, 4,000 to follow by December. ABN Amro is working to 2028, Standard Chartered to 2030, UniCredit on its own schedule. The employment effects will arrive at different speeds in Amsterdam, London, Milan and Frankfurt, and national labour rules will shape how visible they are.
The ECB's survey was conducted before Barclays' October announcement, and it measured intentions rather than outcomes. Its finding stands as a limit on the story: 15% of eurozone firms say cost cutting is why they use AI. The banking sector is currently the clearest exception to that figure.
Does the EU AI Act prevent a bank from replacing staff with AI?
No. The Act regulates how AI systems are built, documented and disclosed, and it bans certain uses such as emotion recognition at work. Redundancy decisions are governed by national labour law, collective agreements and consultation rules, not by the AI Act.
Are the announced cuts final figures?
They are targets. Barclays has already removed about 1,000 roles and plans to remove around 4,000 more by the end of 2026. HSBC, Standard Chartered, ABN Amro and UniCredit have given multi-year targets running to 2028 or 2030, and such targets are regularly revised.
Can an employee ask which decisions about them were made by a machine?
Under GDPR, an employee can request access to personal data an employer holds and can object to decisions based solely on automated processing that produce legal or similarly significant effects. In practice this covers scoring, ranking and profiling, and the employer must be able to explain the logic involved.
