Santander and Nvidia build tools to match AI models to banking tasks
For the CIO at a mid-size US bank
Santander announced on Monday, Oct. 19, that its AI lab is working with Nvidia on capabilities to evaluate AI models and agents for banking, according to PYMNTS, which cited the bank's announcement.1
The work is designed to help determine which AI model is best suited for a task and deploy it efficiently, PYMNTS reported.1 Santander said in a news release provided to PYMNTS that it has been "progressively moving from large general-purpose models towards smaller, specialised models designed for specific banking tasks."1
In selected experiments, the approach cut serving costs from €5 per interaction to approximately €0.03 while upholding strong performance, the release said.1
The AI lab built a "Model Arena" to evaluate open-weight models against banking tasks and decide which model fits each workload, and it is specializing models with the Nvidia NeMo Customizer.1
In one experimental customer-service use case, Santander improved conversation quality while lowering costs by more than 50% by deploying a more efficient AI model, according to the companies.1
Aser Blanco, Nvidia's global banking lead, said: "In this new world, cost per intelligence output is a key metric."1
Santander's figures come from selected experiments, and its one described customer-service case showed a cut of more than 50%, not 99%. Even so, banks weighing AI vendors and model choices may need to track cost per answer by task and test smaller models against their own work.
Santander and Nvidia said they will continue to work together on areas like foundation models, AI agents, fraud and anti-money laundering capabilities, real-time decisioning and advanced simulation technologies, such as banking digital twins, according to PYMNTS.1
Our take. Santander says some tasks fell from €5 to about €0.03 per interaction, roughly 170 times cheaper. Its one detailed customer-service case saved more than 50%. Find out which of those two numbers your AI contracts are priced against.
Action. Vendor management: Pull per-interaction AI serving costs from model contracts and compare each workload against Santander's €5-to-€0.03 result.
What our records say
Santander US's AI strategy is the group's: AI sits inside ONE Transformation, the group programme to run common technology platforms across its markets.1 US-specific AI deployments, budgets and headcount are not disclosed separately.2 At the 25 February 2026 investor day, executive chair Ana Botín said AI would contribute more than €1 billion to the 2028 targets: roughly €300 million of extra revenue and €700 million of lower costs.3
From our records
- Santander USbank · latest Riley: AI lets her 'vibe code' a CRM tool herself (Oct 2026) · vendors FICO · enforcement 2 actions since 2016; latest CFPB consent order, $4.8 million (Dec 2020)