Moving to smaller models reduced Santander's costs by more than 95%

Santander Bank 's AI model-agnostic approach has been successful as the bank deploys specific models for specific tasks for maximum impact and cost savings. The $2.2 trillion bank has cut the cost of handling some customer interactions from roughly €5 ($5.60) to about 3 cents by swapping large general-purpose AI models for open source smaller, task-specific ones, a Santander spokesperson told FinAI News. [caption id="attachment_57372340" align="alignright" width="303"] (Courtesy/Bloomberg)[/caption] The swap reduced costs by more than 95% in selected experimental use cases, achieved “while maintaining strong performance,” the spokesperson said. "We take a model-agnostic approach and select the model best suited to each specific task, balancing performance, cost, scalability, governance and security," the spokesperson said. "Frontier models are used where their capabilities are genuinely required, while smaller specialized models can be more appropriate for specific banking tasks." The bank declined to disclose its split between closed- and open-weight systems. AI value generation The Madrid-based bank reported on Oct. 28 during its third-quarter earnings call that it has generated $217 million in added business value with AI year to date through September. Santander defines business value as incremental revenue, savings and lower loan-loss provisions from its data and AI initiatives. Asked whether the run rate keeps pace with Executive Chair Ana Botin ’s roughly $1.1 billion AI ambition for 2026-28, the spokesperson said the bank is “on track,” noting the added business value compares with an expectation to exceed $224 million for the full year. Botin has targeted about $336 million in added revenue and $784 million in cost reductions from AI by 2028, and the spokesperson said dedicated governance ensures AI benefits are not double-counted against the group’s broader ONE Transformation program. The spokesperson said proof of the bank’s success includes: In Brazil, AI has made filing card fraud claims roughly 95% faster, with up to 90% automation and an error rate below 1%; Its Open banking platform, Openbank, runs about 100,000 anti-money laundering alerts a year through AI; and In the United Kingdome, the bank is routing some 240,000 card-related calls to AI-enabled self-service this year, freeing up around 45,000 employee hours. AI is one lever among several in the efficiency drive. The spokesperson said it should add about one percentage point to the improvement in the group’s cost-to-income ratio by 2028. Santander is targeting a cost-to-income ratio of nearly 36% by 2028, down from 42.6% in the first nine months — already its best level in 17 years, the spokesperson said. The spokesperson framed Santander’s 189 million customers as a “structural AI advantage,” arguing the edge comes from pairing scale with proprietary data and common platforms, “which allows us to build capabilities once and deploy them across markets.” Editor's Note: All amounts have been converted to U.S. dollars. Register here for early-bird pricing for the FinAi Banking Summit, taking place March 1-2 in Denver. View the full event agenda here .

Moving to smaller models reduced Santander’s costs by more than 95%

Bank on track to exceed $224M in AI value generation for 2026

Affirm CEO: Agentic commerce will be messy at first, but AI will learn

Related Posts

Travis Credit Union taps Finalytics.ai for personalized digital experience

Fiserv expects AI implementation to generate $1B in savings

Grasshopper sees Claude, ChatGPT connector queries double monthly

EMERGING FINTECH DIRECTORY

FinAi Podcast

SPONSORED

The Next Generation of Cyberthreats: Powered by AI

Build an Antifragile Strategy to Outperform the Market

How AI and Product Experts Turn Fuzzy Requirements Into Focused Dev-ready Roadmaps

Connect

Welcome Back!

Retrieve your password

Continue Reading with FinAi News Premium - Less than $2/Day