From promise to figure: how Santander quantifies its AI-first strategy

On 22 June 2026, Banco Santander published a progress report that made the industry take notice. Ricardo Martín Manjón, the bank's Chief Data and AI Officer, put a figure on the value of the AI-first strategy with concrete actuals for the first time: 35 million euros in the first quarter of 2026, with the expectation of exceeding 200 million euros by year-end. At the same time, the bank announced it would give all 185,000 employees access to AI tools. For an industry that has debated the return on investment (ROI) of artificial intelligence for years, these are rare figures, precisely because they are concrete.

The strategic target itself is not new. As early as the Investor Day in London on 25 February 2026, Santander had announced that it wanted to generate more than one billion euros of value per year from data and AI by 2028. The June report is therefore not a starting gun but a first interim assessment. And precisely because such figures are so rare, a close look is worthwhile: what do they bear, and what do they leave open?

The numbers Santander presents

The operational core of the report is considerable. By its own account, Santander runs more than 280 automation agents in production, taking over manual tasks in lending, fraud detection, customer identification (Know Your Customer, KYC) and operations. In software development, over 17,000 employees were already working with agentic AI in May; in June, by Santander's measurement, 40 per cent of code was developed with AI support. In Brazil, AI accelerated the handling of credit-card fraud cases by around 95 per cent, with an automation rate of up to 90 per cent and an error rate below one per cent. The models of the digital bank Openbank process around 100,000 anti-money-laundering (AML) alerts per year.

On the value figures, a clarification is worth making, one the June report itself somewhat blurs. The authoritative target comes from the Investor Day: more than one billion euros annually by 2028, not cumulative over three years. With 35 million euros in the first quarter and a full-year target of more than 200 million euros for 2026, the annual run-rate would therefore have to increase roughly fivefold by 2028. That is an ambitious path, not a foregone conclusion.

What AI-first means for Santander

Behind the numbers lies a deliberate method. Santander does not bet on the broadest possible spread of pilot projects but on focus. Technologically, the bank pursues a multi-provider approach: Microsoft Copilot covers everyday productivity, while for specialised applications the models of OpenAI, Anthropic and Google are added, complemented by partners such as the Emirati company G42. This architecture avoids dependence on a single provider, a point that is becoming strategically more important for regulated institutions.

In payments, Santander has also taken a pioneering role: by its own account, the bank was the first in Europe to test payments with AI agents together with Mastercard, and the first in Latin America to do so with Visa. The rollout of AI access to all 185,000 employees is the next step in carrying the strategy from the specialist corner into the breadth of the organisation.

The second look: what the numbers do not say

As concrete as the figures appear, their contextualisation matters just as much. The central caveat: the reported value is not externally audited. It is a corporate communication, not an audited financial report. What exactly feeds into the category of business value is not publicly documented. Avoided fraud losses, saved processing time, additional revenue and reduced error rates are not the same thing in business terms, and it remains open how Santander offsets them.

There is also a blank space on the cost side. Santander discloses no isolated AI investment costs at group level. Without that offset, the value of 35 million euros is not an ROI statement but a gross figure. On top of this, Santander expects more than half of the target value to come from cost savings, an indication of how closely efficiency promises and staffing questions are linked. A further caveat concerns causality: Santander posted a record profit of 14.1 billion euros in 2025, and the group-wide transformation programme has been running for years. How much of the value is genuinely AI-specific and how much would have arisen without AI is hard to separate from the outside. Finally, the showcase examples are granular: the 95 per cent in Brazil applies to a single use case in one market, the 100,000 AML alerts to the digital subsidiary Openbank, not to the group.

In competition: where Santander stands

The comparison puts the figures further into perspective. Singapore's DBS Bank already realised economic AI value of around one billion Singapore dollars in 2025, underpinned by more than 2,000 models and over 370 use cases. Santander's target for 2028 therefore moves in an order of magnitude that a competitor has already reached. The United Kingdom's Lloyds Banking Group also quantifies its AI value contribution, at 50 million pounds for 2025 and a target of more than 100 million pounds for 2026. JPMorgan Chase, in turn, runs over 450 AI use cases but discloses no comparable ROI figures. Santander thus belongs to the leading group, but it is not the lone benchmark the headlines sometimes paint it to be.

One final piece of context belongs here, one Santander itself does not make prominent. Reuters reported on 24 June 2026 that the bank is negotiating voluntary early retirements for up to 3,000 employees in Spain. Santander has not quantified a direct link with the AI strategy; the bank did not disclose how many jobs will be affected. Yet the simultaneity of efficiency promises and staff reductions is the subtext that every AI-first narrative carries with it.

Recommended actions for banks

For institutions watching Santander's path, the value lies less in the absolute numbers than in the method. Five lessons can be drawn.

Christian Schablitzki

Strategy & Management Consultant · Agentic AI expert for financial institutions

More than 20 years in investment banking and derivatives trading, followed by over 10 years advising financial institutions. Currently a Partner at Infosys Consulting in Germany. Certified in Google AI, Generative AI Leader (Google Cloud) and IBM RAG and Agentic AI.

LinkedIn profile →