At a capital markets day in Paris this week, executives at Société Générale did something no large European bank had managed before: they attached a specific euro figure to the cost savings they expect from artificial intelligence and wrote it into the bank’s official financial plan.
The French lender said it expects AI applications to deliver between €500 million and €600 million of cost reductions over the coming years, with roughly €350 million of that already committed through 2029. The announcement, made on September 21, came as the bank signed an agreement with Anthropic to deploy the startup’s Claude model across its operations.
The significance lies less in the purchase itself, since banks have been buying AI software for years, than in the decision to treat the savings as a committed line rather than an aspiration. Analysts said that makes Société Générale the first European bank to embed AI-driven cost cuts into the formal financial planning that investors and regulators hold it to.
The bank’s executives argued that AI can unlock value in several parts of the business at once. It can generate reports automatically, monitor key performance indicators, cut the cost of writing code, and expand the advisory capacity that employees can offer to customers, the bank said in its presentation.
For a lender that has spent years trailing larger rivals such as BNP Paribas and Santander on profitability, the promise of half a billion euros or more in savings carries particular weight. Société Générale has been shrinking its footprint, selling off units in Africa and Eastern Europe, and betting that a leaner, more focused bank can earn more for shareholders than a sprawling one.
Chief Executive Slawomir Krupa, who took the top job in 2023, has made efficiency the central theme of his tenure. Under his watch the bank has exited several businesses and pushed a simpler structure, arguing that scale without discipline is no advantage.
The Anthropic agreement places Société Générale among a growing number of financial institutions experimenting with large language models, though most have stopped short of attaching a specific number to the outcome. The bank did not disclose the financial terms of the deal, and it did not say how quickly it expects the savings to flow through its income statement.
Anthropic, founded by former OpenAI researchers, has positioned Claude as a rival to OpenAI’s models and has been courting corporate customers, including banks, insurers, and law firms. The startup has argued that its models are well suited to the kind of document-heavy, compliance-bound work that defines much of what a bank does.
The claim that a single institution can cut €500 million or more through AI reflects a broader conviction spreading through European finance. Morgan Stanley analysts estimated earlier this year that AI could eventually reduce headcount across the region’s banking industry by as much as one-fifth.
That estimate, which circulated widely through the industry, frames the stakes for employees as clearly as for shareholders. What banks describe as efficiency or advisory capacity can, over time, translate into fewer people doing the same work. Société Générale did not say how many jobs its AI plans might ultimately affect.
European banks have been moving toward this moment for years. They have digitized retail services, closed branches, and relocated back-office work to lower-cost locations. AI is the next step in the same direction, applied this time to white-collar tasks that were long considered beyond the reach of automation.
The French lender is betting that Claude can absorb reporting and coding work that currently occupies thousands of hours of staff time. The €350 million already locked into the plan is the portion the bank feels confident enough to guarantee through 2029; the rest depends on how effectively the tools are deployed and how quickly employees adopt them.
Investors have been waiting for a European bank to demonstrate that AI can move the numbers on an income statement rather than simply appear in strategy presentations. Société Générale’s move is an attempt to do exactly that, and rivals will be watching closely to see whether the savings materialize on schedule.
The risk is that the promised savings arrive more slowly than the plan assumes, or that they come from cutting headcount in ways that weaken the bank’s ability to serve customers over time. Analysts said the coming years will test whether AI is a genuine cost lever or the latest in a long line of efficiency programs announced with more fanfare than follow-through.
What is no longer in question is that the pilot-project phase is ending for Europe’s banks. Société Générale has shifted the discussion from what AI might do to what it is expected to deliver, in euros, on a schedule. That is a different conversation for the industry, and one its competitors will now be pressed to join.
