ISLAMABAD: Pakistan’s rapid growth in digital banking and payments is creating the data and technology base for banks to expand artificial intelligence into treasury operations, including foreign-exchange forecasting, liquidity management and risk monitoring.
Banking professionals say the next phase of financial modernisation could move beyond customer-facing digital services and into core internal functions, where AI tools may help banks analyse transaction flows, anticipate funding needs and improve treasury decision-making.
State Bank of Pakistan data show retail payments reached 3.7 billion transactions worth Rs168.8 trillion during January-March 2026. Digital channels accounted for 92% of transaction volume, handling 3.4 billion payments worth about Rs68 trillion.
Mobile banking and digital-wallet registrations had also crossed 132 million by March, highlighting the scale of Pakistan’s expanding digital financial ecosystem.
Banks Gain Capacity to Invest in Advanced Technology
The sector has also strengthened financially, giving banks more room to invest in advanced systems.
According to the Pakistan Economic Survey 2025-26, banking assets increased 17.8% to Rs63.2 trillion by the end of December 2025, while deposits rose 24.7%.
After-tax profit increased to Rs716 billion from Rs644 billion a year earlier, while the sector’s capital adequacy ratio improved to 20.8%.
Banking professionals say this combination of stronger balance sheets, higher digital transaction volumes and growing data availability could support controlled adoption of AI in treasury, liquidity and risk-management functions.
Global Banks Expand AI Use in Treasury
The shift is already visible internationally.
Ant International recently said Citi, HSBC, Deutsche Bank, Standard Chartered and Barclays were among six major banks partnering on its upgraded Falcon time-series AI model, designed for financial forecasting and liquidity-risk management.
The company said more accurate forecasting could reduce foreign-exchange hedging and allocation costs by more than 60%, although that figure represents Ant International’s own estimate and is not an independently verified industry benchmark.
The broader trend is also reflected in major global banks.
Bank of America said in June that rising complexity in trade and capital flows, foreign-exchange volatility and liquidity risk were increasing demand across Asia-Pacific for AI-supported treasury, trade and currency solutions.
The bank said it spends more than $13 billion annually on technology.
Reuters reported in July that major Wall Street banks were expanding the use of agentic AI into areas including trading and treasury. A KPMG survey cited by Reuters found that 51% of banks were already piloting AI agents.
Pakistan Has Digital Scale, but AI Treasury Use Remains Limited
Pakistan has built substantial digital transaction infrastructure, but official 2025-26 disclosures do not indicate comparable sector-wide deployment of specialised AI models for foreign-exchange forecasting, hedging or intraday liquidity management.
That makes treasury one of the areas where further adoption could emerge.
Foreign-exchange liquidity management remains a central banking function, particularly in an environment where exchange-rate movements, corporate payment flows and funding requirements can change quickly.
SBP data placed Pakistan’s total liquid foreign-exchange reserves at $22.59 billion on August 21, while the weighted-average dollar rate on August 27 stood at Rs277.2321 bid and Rs277.6572 offer.
Banking professionals say AI would not remove currency risk, but could help treasury desks process large volumes of transaction data, identify payment patterns and improve liquidity forecasting.
Experts Favour Human-Supervised AI Models
Muhammad Zafar, Treasury Manager at Soneri Bank, told Wealth Pakistan that Pakistani banks should initially use AI as a decision-support tool rather than move directly towards autonomous foreign-exchange execution.
He said AI models could help forecast intraday dollar demand, identify recurring corporate payment patterns and improve hedge timing, while experienced dealers retained final authority over pricing and execution.
Zafar said effective deployment would depend on reliable historical transaction data, integration with treasury-management systems and rigorous model validation.
He added that pilot programmes should include audit trails, exposure limits and human approval to prevent faster forecasting from creating new model-risk or compliance concerns.
Tahir Awan, Assistant Manager at Wallstreet Exchange Company, said Pakistan’s main challenge was likely to be data architecture rather than access to AI technology.
He said banks had rapidly digitised payments, but treasury, trade-finance and risk information could still be spread across separate systems, limiting the quality of real-time data needed by specialised forecasting models.
Awan said banks and regulators could begin with controlled pilots in liquidity forecasting and foreign-exchange risk management, supported by common standards for data governance, cybersecurity and model accountability.
He said the next phase of banking technology would increasingly depend on how effectively financial institutions use their existing digital infrastructure and transaction data to support faster, better-informed and well-governed decisions.