Summary
Barclays, Citi, Deutsche Bank, and Standard Chartered have integrated the Falcon Time-Series Transformer (TST), a forecasting AI model built by Singapore-based Ant International, into their foreign-exchange systems. Ant International released a new version (FalconTST 2.0) that can lead a public industry benchmark. The model forecasts cash flow and foreign-exchange exposure for cross-border payments. Each bank uses it within its own systems: Barclays in BARX NetFX, Citi with Fixed FX Rates, Standard Chartered alongside SCALE FX under MAS's PathFin.ai program, and Deutsche Bank in its FX operations.
Key Points
- FalconTST (Time-Series Transformer) built by Singapore-based Ant International
- Adopted by Barclays, Citi, Deutsche Bank, Standard Chartered in their FX systems
- Barclays: BARX NetFX platform; Citi: Fixed FX Rates; Standard Chartered: SCALE FX under MAS PathFin.ai; Deutsche Bank: FX operations
- FalconTST 2.0 recorded MASE of 0.666 on a leading public benchmark for time-series foundational models
- Top of the leaderboard ahead of similar models; forecast accuracy above 93%
- First deployed internally at Ant International to manage its own cash flow and currency exposure
- Time-series models predict transaction volumes, account balances, currency positions
- Forecasts feed into FX hedging; inaccurate predictions can lead to over- or under-hedging
- Plans to make the model a reusable forecasting tool across finance, aviation, e-commerce, logistics
- Ant International chief innovation officer Jiang-Ming Yang: value is turning forecasts into decisions on liquidity, currency exposure and capital allocation
- Kelvin Li (GM, platform technology): improved accuracy in 2.0 extends benefits to banking partners and sectors like e-commerce, travel and fintech
- Singapore is home to Ant International, which is part of the Ant Group ecosystem