Artificial intelligence in Basel's banks

Read the PDF edition (3 pages)

The banks of Basel range from the branches of UBS through the Basler Kantonalbank and its subsidiary Bank Cler to private banks and the local offices of foreign institutions, as Part E, Chapter I described. Their use of artificial intelligence follows the pattern of the Swiss banking industry as a whole, with the largest institutions running dedicated programmes and smaller banks relying on the products of software suppliers and core banking providers.¹ A FINMA survey published in 2024 found that around half of the institutions questioned already used artificial intelligence or were developing applications, with an average of several applications per institution. The survey also found that most of these applications drew on models supplied by a small number of external providers.²

Anti-money laundering is the field where the gains are most tangible. Traditional transaction monitoring relies on rules, such as thresholds for cash deposits or alerts for transfers to high-risk countries, which generate large numbers of false positives that analysts must review one by one. Machine learning models can rank alerts by risk, detect networks of related accounts and identify patterns that no fixed rule anticipated. The Anti-Money Laundering Act and the Anti-Money Laundering Ordinance of FINMA leave the choice of method to the institution, provided that its monitoring system is effective and its clarifications are documented.³ The same tools help with know-your-customer processes. They extract data from passports, corporate registers and trust deeds, check names against sanctions lists and adverse media and assemble the profile that the compliance officer reviews before opening a relationship.

Fraud prevention is a second established field. Card issuers and payment service providers in Switzerland use real-time models to score transactions, while banks use behavioural analytics to detect account takeover and authorised push payment fraud, in which clients are manipulated into transferring money to criminals. Credit decisions form a third field, although Swiss banks rely on automated scoring mainly in consumer credit and leasing and much less in mortgage lending and lending to small firms, where relationship managers still decide on the basis of personal knowledge and collateral.⁴

Advisory services have been slower to change. Robo-advisers that construct portfolios from questionnaires appeared in Switzerland from about 2015, offered by start-ups, cantonal banks and the large banks, yet they have remained a modest share of the market for wealth management. Generative models have opened a different route. Relationship managers now use internal assistants that summarise research, prepare meeting notes, draft client letters and check whether a recommendation is consistent with the client's investment profile as recorded under the Financial Services Act.⁵ In private banking, where Basel has a long tradition, the technology supports the adviser without replacing the personal relationship that wealthy families still expect.

Client communication and software development complete the picture. Chat assistants answer routine questions on the bank's website and in its app, while voice systems route calls and authenticate callers. Inside the bank, developers use code assistants to write and test software, which matters because Swiss banks maintain large and often old core banking systems. Several institutions have reported productivity gains in software development that exceed those in any other function, although published figures vary widely and depend on how productivity is measured.⁶

Notes

1. On the Basel banking landscape, see Part E, Chapter I. On the use of AI by Swiss banks of different sizes, see the surveys of the Swiss Bankers Association and FINMA's supervisory communications since 2023. Status to be verified. ↩
2. FINMA, Risk Monitor 2024 and the accompanying communication on the survey of AI use among supervised institutions, published in spring 2025. Figures and dates to be verified. ↩
3. Arts. 6 and 7 of the Anti-Money Laundering Act (AMLA, SR 955.0); Art. 20 of the FINMA Anti-Money Laundering Ordinance (AMLO-FINMA, SR 955.033.0) on the computer-based monitoring of transactions. See Part E, Chapter III. To be verified. ↩
4. On fraud detection and credit scoring in Swiss banking, see the annual reports of the Swiss Bankers Association and the literature on the Consumer Credit Act (CCA, SR 221.214.1), which requires a creditworthiness assessment under Arts. 28 to 30. To be verified. ↩
5. On suitability and appropriateness checks under Arts. 10 to 14 of the Financial Services Act (FinSA, SR 950.1), see Part E, Chapter II. On robo-advice in Switzerland, see the studies of the Lucerne University of Applied Sciences and Arts (IFZ) on digital wealth management. To be verified. ↩
6. On the use of code assistants and productivity measurement in financial institutions, see BIS, Annual Economic Report 2024, Chapter III, and the studies cited there on generative AI and productivity. Figures to be verified. ↩

For general information only. This article does not constitute legal or tax advice, does not replace advice on an individual case and creates no lawyer-client relationship. It reflects the legal position as of August 2025.