Switzerland's Incore Bank has completed a proof-of-concept project deploying agentic artificial intelligence to accelerate digital customer onboarding and automate substantial portions of its risk-assessment workflow — a development that signals a meaningful shift in how mid-tier European banks are beginning to operationalize the latest generation of AI tools. The trial, executed in partnership with technology services firm Kyndryl and Google Cloud, harnessed Google's Gemini large language models alongside Kyndryl's proprietary agentic AI framework, positioning the Swiss institution at the forefront of a rapidly evolving competitive landscape in digital banking infrastructure.
Beyond Chatbots: The Rise of Agentic AI in Banking Operations
The distinction between conventional AI assistants and agentic AI is not semantic — it is architectural. Where earlier generations of banking AI functioned as reactive tools, answering queries or flagging anomalies when prompted, agentic systems are designed to pursue multi-step objectives autonomously. They can gather data from disparate sources, reason across that information, execute sequences of decisions, and course-correct without requiring human intervention at every juncture. In the context of customer onboarding and risk assessment — processes historically characterized by manual document review, compliance handoffs, and sequential approval chains — this autonomous capability carries substantial operational implications. Incore Bank's decision to test this architecture in a structured proof-of-concept reflects a growing recognition across the European banking sector that the next frontier of efficiency gains will not come from digitizing paper forms, but from allowing AI agents to own entire workflow segments end-to-end.
What the Proof-of-Concept Targeted
The specific ambitions of the Incore Bank trial centered on two of the most resource-intensive phases of the client lifecycle in institutional and digital banking. First, digital onboarding — the process by which a new customer is verified, credentialed, and brought onto the bank's books — has long been a friction-heavy experience both for clients and for the compliance teams responsible for Know Your Customer and Anti-Money Laundering checks. Second, risk-assessment automation sought to reduce the human analyst burden associated with evaluating counterparty exposure, credit profiles, and regulatory risk flags. By deploying Gemini models within Kyndryl's agentic framework, the PoC explored whether AI agents could navigate these tasks with sufficient accuracy and regulatory coherence to warrant broader implementation. The combination of Google Cloud's infrastructure scale with Kyndryl's enterprise integration expertise provided a technically credible environment in which to stress-test these assumptions.
Why This Collaboration Architecture Matters
The three-party structure of this initiative — a regulated Swiss bank, a global technology services integrator, and a hyperscale cloud provider — is itself instructive. Incore Bank brings the regulatory context and client-facing requirements; Kyndryl provides the enterprise AI orchestration layer and the implementation muscle to connect agentic systems to existing banking infrastructure; Google Cloud supplies the foundational model capability through Gemini and the compute backbone to run it at production scale. This division of labour reflects how banks are increasingly approaching AI adoption: not by building from scratch internally, nor by simply subscribing to a software-as-a-service product, but by assembling bespoke capability stacks from specialist partners. The arrangement also distributes regulatory and operational risk across parties with clearly delineated responsibilities — a factor of no small consequence in Switzerland's stringent financial supervisory environment under FINMA oversight.
Switzerland as a Testbed for Regulated AI in Finance
Switzerland's positioning as a pilot environment for this class of technology is no accident. The country's financial regulator, the Swiss Financial Market Supervisory Authority (FINMA), has maintained a relatively principles-based approach to technology adoption, providing institutions with interpretive room to experiment while retaining clear accountability expectations. Swiss banks have historically invested heavily in back-office automation and wealth management technology, and the country's dense concentration of private banking, asset management, and banking-as-a-service providers creates a competitive environment where operational efficiency directly translates into margin preservation. Incore Bank, which operates within Switzerland's banking-as-a-service and digital banking infrastructure space, occupies a position where onboarding speed and compliance automation are not peripheral concerns — they are core to the value proposition offered to downstream clients and partner institutions.
The Compliance Dimension: Automating Risk Without Abdicating Responsibility
Automating risk checks is, perhaps, the most consequential and most scrutinized application of AI in regulated banking. Errors in Know Your Customer verification or Anti-Money Laundering screening carry not just financial penalties but reputational and criminal liability. Agentic AI systems must therefore operate within tightly defined parameters, with human oversight mechanisms embedded at critical decision nodes. The proof-of-concept format that Incore Bank chose is precisely the appropriate methodological framework for this domain: contained, measurable, and designed to surface failure modes before any system touches live customer data at scale. The involvement of Kyndryl — a firm with deep enterprise risk and compliance integration experience — suggests that the architecture prioritized auditability and governance alongside raw automation capability.
What This Means for European Digital Banking
Incore Bank's completed proof-of-concept, while a single data point, reflects a broader directional shift in how European banks are beginning to treat agentic AI — not as a distant strategic aspiration but as a near-term operational tool worthy of structured evaluation. As the European Banking Authority and national regulators continue to develop guidance around AI model risk and algorithmic accountability, institutions that have already accumulated proof-of-concept evidence — including documented performance, failure modes, and governance frameworks — will be better positioned to seek regulatory approval for live deployments. The partnership between Incore Bank, Kyndryl, and Google Cloud demonstrates that the infrastructure for agentic AI in banking is available today. The remaining questions are ones of governance, scale, and the speed at which regulatory frameworks can mature to match the pace of technical ambition.
Written by the editorial team — independent journalism powered by Codego Press.