Incore Bank’s AI Hits 99% Accuracy on KYC Checks
Know-your-customer (KYC) checks are expensive partly because a bank isn’t simply verifying an identity. Compliance teams have to collect information from documents, reconcile it across systems, investigate potential risk factors and build an auditable record explaining how they reached a decision. Incore Bank, a Swiss B2B transaction bank, is testing how much of that work […] The post Incore…
A Swiss B2B transaction bank, Incore Bank, is exploring the potential of artificial intelligence to streamline and automate Know-Your-Customer (KYC) checks. KYC procedures are complex, requiring the collection of information from various documents, reconciliation across systems, investigation of risk factors, and the creation of an auditable record.
Incore Bank's test, built using Kyndryl's Agentic AI Framework and Google Cloud's Gemini models, achieved up to a 99% accuracy rate in extracting data from onboarding documents. The AI agents divided the tasks, with some focusing on extracting and verifying customer information, others gathering supporting evidence, identifying risk factors, and generating an explainable risk score.
While the 99% accuracy figure showcases the system's efficiency in pulling relevant data from documents, it does not imply the AI made KYC decisions correctly 99% of the time. This specific performance metric emphasizes the importance of human oversight in the final risk decision-making process. At a time when financial firms are heavily investing in AI, Incore Bank's experiment highlights the limited adoption of AI in KYC processes.
Only 20% of financial services and insurance firms currently use AI for KYC, know-your-business checks, and identity verification. However, 65% of firms plan to adopt or expand AI in these areas within the next 12 months, indicating a rapid shift in the industry. Incore Bank's AI agents aim to automate the evidence-gathering phase of KYC, reducing the time spent on this task while leaving the actual decision-making and accountability to human compliance officers.
Written by urgent.news from PYMNTS's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
