The share of National Bank of Kazakhstan personnel utilizing AI services on a monthly basis rose from 5–6% in early 2026 to over 80%. The financial regulator is currently preparing a pilot of specialized software agents engineered to analyze digital wallet flows and flag suspicious transactions.
The operational rollout was detailed by Deputy Chairman of the National Bank Binur Zhalenov in an interview with Digital Business published October 5. According to Zhalenov, adoption metrics reflect employee retention and repeated tool usage following initial trials.
AI Generates Analytical Drafts, Humans Retain Final Decisions
For citizens, the public-facing assistant “Fraudie” operates on antifraud.kz, explaining remediation steps in cases of financial fraud. In parallel, an investigator AI assistant parses incident documentation and compiles draft decision memoranda for professional officer review.
A pilot deployment of autonomous AI agents in financial supervision is scheduled for Q4 2026. The architecture is engineered to isolate anomalous financial transactions and prepare briefing packets for human analysts, including units within the National Cryptocurrency Analytics Center. Final regulatory authority, Zhalenov emphasized, rests strictly with human officers.
Data Foundations Preceded Production Rollouts
This initiative represents the sequential advancement of a previously announced strategic roadmap. In an official release dated January 29, the National Bank confirmed the deployment of an enterprise Data Factory and a proprietary GPU computing cluster. A strategic priority for 2026 has been scaling the AISANA multi-agent AI framework natively wired to the regulator’s internal data stores.
This engineering sequence is critical for contextualizing the announcement: prior to delegating analytics to automated models, an institution must aggregate harmonized data pipelines and establish rigorous governance access tiers. Rising internal adoption indicates tool dissemination within the central bank; it does not independently quantify recommendation accuracy or supervisory efficacy.
In the interview, Zhalenov reported more than ten distinct AI solutions currently deployed in production. For workflows involving confidential banking data, models are strictly hosted on on-premises sovereign infrastructure.
The 134-Agent Metric Pertains to Freedom Bank
A concurrent Digital Business report from the forum referenced over 134 operational AI agents. This figure strictly applies to the AI Compliance framework deployed at Freedom Bank, where agents audit transaction flows and customer behavioral patterns. Attributing this metric to the central bank would be inaccurate.
During the same panel, National Bank representatives detailed an experimental monetary policy advisor. The “Dove” persona formulates empirical arguments supporting policy easing, the “Hawk” models arguments for monetary tightening, and the “Owl” synthesizes comparative trade-offs. This operates as an analytical brief generator; benchmark interest rate decisions remain the statutory prerogative of the Monetary Policy Committee.
RAEM.KZ previously reported on enterprise banking AI adoption during AWS Cloud & AI Day. The National Bank’s latest disclosure extends this technological arc from the regulatory apex: algorithmic tools assume expanding analytical preparation duties, while statutory governance remains firmly with human leadership.
