Halyk Bank: Automating Currency Payment Authorization at Kazakhstan's Largest Bank
25% of outgoing currency payments authorized automatically with precise error flagging — streamlining treasury operations at a $45B institution.
About the Client
Halyk Bank is the largest financial institution in Kazakhstan and Central Asia, with total assets exceeding KZT 20.9 trillion (approximately USD 45 billion). Founded in 1923 and listed on the Kazakhstan Stock Exchange, the London Stock Exchange, and the Astana International Exchange, the bank holds the highest credit rating among private banks in Kazakhstan (Fitch BBB-, Moody's Baa1) and operates over 530 branches across the country, with additional presence in Georgia and Uzbekistan.
The Challenge
Manual Currency Payment Authorization at Massive Scale
As one of Kazakhstan's systemically important financial institutions, Halyk Bank processes an enormous volume of outgoing currency payments daily — cross-border transfers, trade finance settlements, and foreign-currency transactions that each require validation against regulatory requirements, counterparty details, and internal compliance rules before authorization.
These checks are deterministic: each payment must pass through a consistent set of rules covering amounts, beneficiary data, purpose codes, and documentary compliance. Yet manually reviewing every outgoing currency payment consumed significant treasury staff capacity.
Every transaction required a human operator to verify data fields, cross-check against compliance rules, and either authorize the payment or flag it for investigation. At Halyk Bank's transaction volumes, this created a persistent bottleneck — one where the majority of payments were routine and compliant, yet each still required the same level of manual attention.
The Solution
Automated Payment Authorization with Human-in-the-Loop
Halyk Bank implemented a Menelaus AI robot to automate the authorization of outgoing currency payments in the treasury function — targeting the clearly compliant population while ensuring payments with potential issues receive human judgment.
Rule-Based Payment Authorization
The bot applies the bank's full validation ruleset to each outgoing currency payment — checking beneficiary details, transaction amounts, purpose codes, and compliance parameters. Payments that pass all checks are authorized automatically, without human intervention.
Intelligent Error Flagging
For payments that do not pass validation, the bot identifies the specific location and nature of the suspected error — pinpointing exactly which field or rule triggered the exception. This transforms the reviewer's task from a full re-examination to a targeted review of a specific flagged issue.
Selective Automation at Scale
The deployment achieved automatic authorization of 25% of all outgoing currency payments. The remaining 75% are flagged with precise error locations for human review — a deliberate human-in-the-loop design that financial regulators expect in payment authorization.
The 25/75 Split
The 25% automatic / 75% human-reviewed split reflects a deliberate design choice for regulated banking: automation targets the clearly compliant population while ensuring any payment with potential issues receives human judgment — exactly the human-in-the-loop model that financial regulators expect. Each flagged payment comes with a specific indicator of where the potential error exists, converting full-scope manual review into targeted exception handling.
The Results
Streamlined Treasury at Kazakhstan's Largest Bank
The 25% automatic authorization rate eliminates manual processing for a quarter of the bank's outgoing currency payment volume — a meaningful throughput gain at Halyk Bank's scale. Equally important, the remaining 75% of payments come with specific flags indicating where potential errors exist, converting full-scope manual review into targeted exception handling.
At the transaction volumes of Kazakhstan's largest bank, automating even a quarter of currency payment authorization — and accelerating human review of the rest through precise error flagging — represents a substantial reduction in treasury processing time and a measurable improvement in throughput.
Why It Matters
This early-generation deployment at Halyk Bank established an important precedent: RPA can operate within the most sensitive layer of banking operations — payment authorization — when designed with the right governance model. The 25/75 split demonstrates that automation in regulated banking does not require an all-or-nothing approach.
By starting with the clearly compliant subset and providing precise error flags for the rest, the solution respects the regulatory requirement for human oversight while eliminating unnecessary manual effort. As an early Menelaus AI deployment in Kazakhstan's banking sector, this project laid the groundwork for the platform's subsequent adoption by other major financial institutions in the region.
"We have made a robot that automatically authorizes 25% of outgoing currency payments. In other cases, if there is a possible error, it notes exactly where it may be."
Ready to automate payment operations?
Talk to an automation specialist who knows banking — and your compliance requirements.