CONCRYT deploys agentic AI to monitor payments

CONCRYT has installed agentic artificial intelligence in its live payment gateway to monitor transaction flows and recommend routing adjustments. The system has processed all of the company’s payment traffic for the past eight months and covers more than three million transactions.

The monitoring tool enforces roughly 100 automated rules, with some checks running as frequently as once per minute. It monitors payment traffic, approval rates, financial performance, operational limits and merchant behavior. The system raises alerts when indicators shift unexpectedly and provides recommendations on where volume could be re-routed; operations teams retain final authority over any changes.

CONCRYT’s operations team applies the tool daily to adjust routing with the aim of improving approval rates. So far this year the team has had more than 2,000 interactions with the AI assistant through the company admin panel, Slack and Telegram, covering traffic anomalies, payment performance and individual transactions. The monitoring layer also executes agreed recurring checks without new prompts from staff.

Earlier this year CONCRYT added large language model analytics to its gateway that let merchants query payment data and receive real-time insights via the admin panel and messaging apps. The agentic layer extends those capabilities by analyzing flows, correlating signals and recommending specific routing changes instead of only answering questions.

A 2026 report from the Cambridge Centre for Alternative Finance found that 52% of firms in financial services are actively adopting agentic AI. The report showed fintech firms reporting 57% adoption versus 45% at traditional institutions, and 23% of respondents had moved deployments beyond pilot stages.

“A payment gateway produces more signals every day than any operations team could realistically monitor by hand,” Andrii Shevchuk, CONCRYT’s CTO and co-partner, observed. “By combining continuous monitoring with AI-assisted investigation, our team can move much faster from spotting a change to understanding what is causing it. The system can analyze traffic, identify a drop in approval rates and recommend where volume could be moved, while the final decision remains with our people.”