Incore Bank has completed a test of artificial intelligence systems designed to handle much of the work involved in verifying new customers, achieving accuracy rates that the Swiss B2B transaction bank says could significantly compress a process that currently stretches across weeks or months.
The proof of concept, built with technology partners Kyndryl and Google Cloud, used multiple AI agents to extract and verify information from onboarding documents. The system reached up to 99% accuracy in pulling data from those materials, according to a company announcement describing the test results.
What the accuracy figure actually measures
The 99% metric refers specifically to how well the system extracted information from documents, not whether it made correct judgments about a customer’s risk profile. The distinction matters. Document extraction is one input among several that feed into a compliance officer’s final determination. The AI handles the evidence-gathering step; a human still makes the actual decision and carries the accountability that comes with it.
“At Incore Bank, we believe innovation must go hand in hand with trust, transparency and strong regulatory governance,” said Mark Dambacher, CEO of Incore Bank.
Related: AI Assistants Enter Mainstream Retail Investing
The architecture underlying the test relies on Kyndryl’s Agentic AI Framework working with Google’s Gemini models. Rather than placing the entire task on a single system, the setup coordinates separate agents, each assigned to a distinct piece of the workflow: extracting customer information, cross-referencing external sources, flagging potential risk factors, and generating a risk score along with documentation that compliance staff can review later.
The broader adoption picture
Financial firms are increasing AI spending, but KYC automation remains an outlier in terms of actual deployment. Research from PYMNTS Intelligence found that 85% of financial services and insurance companies plan to raise AI budgets over the next year, with productivity gains cited as a primary driver by 65% of those firms. Yet only 20% currently use AI for know-your-customer checks, identity verification, or know-your-business reviews—the second-lowest adoption rate among the tasks tracked in the research, ahead of only A/B testing.
The gap between planned investment and current use suggests substantial room for systems like the one Incore tested. Sixty-five percent of firms surveyed said they intend to adopt or expand automation for identity verification and KYC within the next twelve months. That figure rises to 90% when longer-term plans are included.
Given that the technology can handle the more time-intensive portions of the compliance workflow, the appeal is straightforward: reducing the manual effort required to gather and reconcile customer data leaves officers more time for judgment calls that actually require human discretion. Whether that tradeoff holds as the systems scale across different customer bases and regulatory environments remains to be seen, but the direction of travel for the industry seems clear.
Related: Vista CFO Favors Jet Subscription Pricing
Why this part of compliance is expensive
KYC checks involve more than confirming an identity matches a government-issued document. Compliance teams must pull information from multiple sources, reconcile data across internal systems, investigate potential red flags, and produce a record that explains how they reached their conclusions. The process generates audit trails regulators can examine later, which adds another layer of documentation requirements.
Automating those steps could cut weeks from timelines that currently extend months. The Incore test suggests the technology is at a point where that outcome is plausible, at least for the data extraction and evidence-gathering components. What remains less certain is how regulators in different jurisdictions will view the shift in responsibility—human in the loop does not mean human in control of every step.
Incore Bank’s announcement frames the work as a proof of concept rather than a deployed product. The bank is testing how much of the compliance workload software can handle before a person needs to step in. Whether other financial institutions move toward similar configurations will likely depend on how the first wave of implementations performs under real-world scrutiny.
