Kyndryl, Incore Bank and Google Cloud Advance Agentic AI for Faster KYC Onboarding
Kyndryl, Incore Bank, and Google Cloud have completed a proof of concept showing how agentic AI can help banks accelerate customer onboarding while keeping risk decisions explainable, auditable, and subject to human oversight.
ZURICH, August 31, 2026 — Kyndryl has partnered with Swiss transaction bank Incore Bank and Google Cloud to demonstrate how agentic artificial intelligence can transform Know Your Customer (KYC) processes in banking.
The proof of concept uses Google’s Gemini models alongside Kyndryl’s Agentic AI Framework to automate several stages of customer onboarding and risk assessment. The approach is designed to reduce the amount of manual work required by compliance teams without removing human control from critical decisions.
For banks, KYC remains one of the most important parts of customer onboarding. Teams must collect and verify information, assess potential risks, and maintain detailed records for regulatory purposes.
These processes can involve large volumes of documents and data from different sources. Manual reviews and repeated handoffs can therefore make onboarding slow and resource-intensive.
The Kyndryl, Incore Bank, and Google Cloud proof of concept aims to address that challenge through a network of AI agents.
Agentic AI Targets Complex KYC Workflows
The solution brings multiple AI agents together to handle different parts of the KYC process.
The agents can work with both structured and unstructured information. They support tasks such as extracting customer data, validating information, gathering evidence and identifying potential risk factors.
The system can also generate explainable risk scores and create auditable records that compliance professionals can review.
A semantic layer, supported by Kyndryl’s policy-as-code capability, helps translate business and regulatory requirements into operational rules. Guardrail agents are also used to support controlled AI decision-making.
This architecture is intended to give banks more automation while maintaining transparency around how risk assessments are produced.
Up to 99% Accuracy in Data Extraction
During the proof of concept, the solution achieved up to 99% accuracy in extracting data from customer onboarding documents, according to Kyndryl.
The project also demonstrated the potential to reduce onboarding timelines from months to days.
For financial institutions, that could have a meaningful operational impact. Faster processing can improve the customer experience while allowing compliance professionals to focus more of their time on cases that require human judgment.
Rather than replacing compliance teams, the model is designed to shift their role toward higher-value decisions and exception handling.
Human Oversight Remains Central
Automation in financial services comes with a major challenge: speed cannot come at the expense of regulatory control.
Kyndryl and Incore Bank have therefore emphasized explainability, auditability, and human governance throughout the proof of concept.
The system continuously evaluates customer information and can support both initial onboarding and ongoing reviews. When circumstances change, AI agents can help identify new information that may affect a customer’s risk profile.
Compliance professionals remain involved in the process, particularly when decisions require judgment or additional investigation.
Mark Dambacher, CEO of Incore Bank, said the project demonstrates how agentic AI can improve efficiency while maintaining the controls expected in a regulated banking environment.
The approach could also help banks achieve greater consistency in risk assessments by applying defined policies across large volumes of customer cases.
Policy-as-Code Brings Rules Into the AI Workflow
One of the notable elements of the project is Kyndryl’s use of policy-as-code.
The technology helps convert policies and business requirements into rules that can be incorporated into automated workflows. This provides an additional control layer for AI agents operating within sensitive financial processes.
For banks, this is particularly important because KYC decisions must be traceable. Institutions need to understand not only the outcome of a risk assessment but also the evidence and reasoning behind it.
By combining AI agents with defined policies and guardrails, the solution aims to create a more controlled environment for deploying generative and agentic AI in compliance operations.
What Agentic AI Could Mean for Banking
The project reflects a broader shift in financial services from AI tools that simply analyze information toward AI systems capable of coordinating multi-step workflows.
Traditional AI applications may flag a suspicious transaction or identify missing information. Agentic AI can go further by coordinating several actions across systems, gathering evidence and preparing an outcome for human review.
That capability could be particularly valuable in KYC, AML, fraud prevention and other compliance-heavy processes.
However, banking institutions still need strong controls around data access, model behavior, security and accountability. The Incore Bank project highlights how those safeguards can be incorporated into an agentic AI architecture rather than added only after deployment.
A Potential New Model for Digital Onboarding
The proof of concept does not represent a full-scale production deployment. Instead, it demonstrates how banks could use agentic AI to redesign parts of the customer lifecycle.
For Incore Bank, the work provides a potential path toward more efficient customer risk management. For Kyndryl and Google Cloud, it also demonstrates how enterprise AI infrastructure can be applied to highly regulated financial workflows.
As banks face growing compliance workloads and increasing customer expectations for faster digital onboarding, the ability to automate routine processes while preserving human oversight is likely to become increasingly important.
The key question is no longer simply whether AI can analyze customer information. It is whether AI can participate in regulated workflows while remaining transparent, controlled and accountable.
The Kyndryl, Incore Bank and Google Cloud proof of concept suggests that agentic AI could be moving closer to that reality.
Kyndryl and Incore Bank collaborate with Google Cloud to help advance agentic AI customer onboarding


Comments
No comments yet. Be the first to share your thoughts.
Leave a comment