Kyndryl, Incore Bank, and Google Cloud have completed a proof of concept applying agentic AI to bank customer onboarding and risk assessment.
The initiative, based on Google's Gemini models, was designed to test whether agentic AI can support Know Your Customer (KYC) processes while keeping decision-making explainable and subject to human oversight.
Switzerland-based Incore Bank worked with Kyndryl, a US-listed enterprise technology services provider, and Google Cloud to explore how automation could reduce the manual workload involved in gathering, verifying, and reconciling customer data across multiple documents and systems. KYC processes are a regulatory requirement for banks, and manual handoffs between teams have traditionally slowed onboarding while adding to compliance workloads.
Agentic AI framework and policy as code
The proof-of-concept applied Kyndryl's Agentic AI Framework together with Google's Gemini models, introducing a semantic layer built on Kyndryl's policy as code capability and a set of guardrail agents. The setup was intended to support end-to-end customer risk assessments, covering both onboarding and ongoing reviews, while monitoring for changes in a customer's circumstances over time.
At the same time, the solution combines several AI agents operating across structured and unstructured data, internal banking systems, and external data sources. According to the companies, these agents support the extraction and validation of customer information, evidence gathering, identification of risk factors, explainable risk scoring, and the creation of auditable decision records intended for use by compliance teams.
During the proof of concept, the solution reportedly achieved up to 99% accuracy in the automated extraction of data from customer onboarding documentation. The companies stated that the approach demonstrated potential to reduce onboarding timelines from months to days, while allowing compliance staff to focus more time on cases requiring judgment, rather than repetitive data-handling tasks. Transparency, control, and human oversight were maintained throughout the process, according to the companies involved.
Regulatory context
Banks operate in a heavily regulated environment where KYC obligations are tied to anti-money laundering and broader risk management requirements. Mark Dambacher, CEO at Incore Bank, said the proof of concept illustrated how cloud-enabled agentic AI could help assess prospective and existing customers more efficiently, while maintaining the oversight expected in a regulated setting.
Moreover, Jacqueline Wild, Managing Director and Vice President, Kyndryl Alps, described agentic AI as a further step in the development of enterprise AI, noting that its application in banking depends on governance, explainability, auditability, secure data access, and human oversight. The companies said the project points to how agentic AI frameworks combined with policy as code capabilities could be applied to modernise onboarding and risk-assessment processes across the customer lifecycle, subject to further development beyond the proof-of-concept stage.
