Incore Bank is a Swiss private bank that has completed a proof of concept with Kyndryl and Google Cloud to test how agentic artificial intelligence (AI) can support the onboarding of new customers and the Know Your Customer (KYC) process.
The proof of concept uses Google Cloud’s Gemini models together with Kyndryl’s Agentic AI Framework to help automate customer information checks and risk assessments.
The project aims to make the KYC process more efficient while keeping decisions explainable and auditable. Human oversight is still included throughout the process.
In usual KYC processes without AI assistance, compliance teams need to collect, verify, and compare information from different documents, systems, and data sources. These tasks can take a significant amount of time and can require staff to perform repetitive manual work.
Behind the Mechanism
The combination of Kyndryl’s Agentic AI Framework with Google’s Gemini models forms the basis of the proof of concept. It also uses a semantic layer based on Kyndryl’s policy-as-code capability and guardrail agents.
The setup is designed to support risk assessments from the beginning of customer onboarding and ongoing customer reviews. It can also monitor changes in a customer’s circumstances over time, which can help banks continue assessing customer risk after the initial onboarding process.
Several AI agents work across structured and unstructured data through internal banking systems and trusted external sources. According to the companies, these agents can extract and validate customer information, gather supporting evidence, and identify possible risk factors.
The system can also produce explainable risk scores or assessments and create auditable records of decisions. These records can then support compliance teams when reviewing how a risk assessment was made.
AI Used Right for Accuracy
During the testing, the proof of concept reached up to 99% accuracy when automatically extracting information from onboarding documents. The companies also said that the solution showed potential to reduce the time needed for customer onboarding from months to days.
The use of AI agents can help reduce the amount of repetitive work originally handled by compliance professionals. This can give compliance staff more time to focus on cases that require human judgment and more detailed review instead of spending most of their time gathering and checking information.
The system is designed to work with different types of information and systems, which can help in collecting the customer information needed for risk assessments. This could make the process more consistent while still allowing compliance professionals to review the results.
However, the proof of concept is still a test of technology. The results show its potential, but this does not mean that agentic AI can fully replace compliance staff or human decision-making in banking.
Human Oversight and Regulatory Governance
A key part of the project is still having human oversight. Since banks operate in a highly regulated environment, the use of AI needs to include controls that allow decisions to be explained, reviewed, and audited.
The CEO of Incore Bank stated that the project shows how cloud-enabled agentic AI can help banks assess both potential and existing customers more efficiently while maintaining the oversight and controls needed in a regulated environment.
The Managing Director and Vice President of Kyndryl Alps also said that the use of agentic AI in highly regulated industries like banking depends on strong governance, explainability, auditability, secure access to data and human oversight.
The project shows how banks are exploring AI to modernize important compliance processes while still maintaining control over how customer risk decisions are made.
Despite the continued use of AI in financial institutions, human oversight remains important to make sure the technology is used responsibly and in accordance with regulatory requirements.
