Incore Bank has completed a proof-of-concept project that explores how agentic artificial intelligence can speed digital customer onboarding while automating large parts of the risk-assessment process.
The trial, conducted with technology services firm Kyndryl and Google Cloud, used Google’s Gemini models together with Kyndryl’s agentic AI framework to handle Know Your Customer (KYC) tasks that traditionally consume weeks or months of manual work.
Banks must collect, verify and reconcile information from many documents, internal systems and external sources before they can accept a new client.
Those steps also have to produce a complete, auditable trail that regulators can review later.
When the work is done by people handing files from one desk to another, the process slows and compliance teams spend most of their time on repetitive checks rather than on cases that require professional judgment.
The pilot solution coordinates several specialized AI agents.
Some agents pull data from unstructured documents and structured records, then check that the information is consistent.
Others search internal databases and trusted external sources for supporting evidence, flag potential risk factors, generate an explainable risk score and compile a decision file that a human compliance officer can review.
A semantic layer and “policy-as-code” rules sit on top of the agents, so the system stays aligned with the bank’s own policies and regulatory requirements.
Guardrail agents further constrain what the models can do.In the test environment the system extracted data from onboarding documents with up to 99 percent accuracy.
Organizers said the same approach could compress an onboarding cycle that now lasts months into a matter of days.
Compliance staff would then concentrate on the minority of files that still need human interpretation, while the technology continues to watch for changes in a customer’s circumstances after the account is opened.
Incore Bank CEO Mark Dambacher emphasized that any new technology must still meet the standards of trust, transparency and governance that Swiss banking regulation demands.
He described the pilot as evidence that cloud-based agentic AI can help banks evaluate both prospective and existing clients more quickly without sacrificing the controls required in a tightly supervised industry.
Jacqueline Wild, managing director at Kyndryl Alps, called agentic AI the next stage of enterprise automation because it can orchestrate entire workflows rather than isolated tasks.
She noted that success in banking depends on explainability, auditability, secure data handling and continued human oversight—qualities the framework and policy-as-code layer were designed to provide.
Incore Bank, a FINMA-regulated B2B transaction bank based in Schlieren, specializes in serving other financial institutions and corporate clients with both traditional and digital-asset services.
The proof of concept is presented as an early illustration of how the bank intends to combine technological innovation with the rigorous risk management expected of a Swiss institution.
The project does not claim that machines will replace compliance officers. Final risk ratings and onboarding decisions remain with people.
What the trial does suggest is that a well-governed set of cooperating AI agents can take over the most time-consuming parts of document review and evidence gathering, leaving specialists free to apply judgment where it matters most.
