Citi, HSBC and Standard Chartered Adopt Ant International’s New AI-Powered Foreign-Exchange Tool

Citi, HSBC and Standard Chartered Adopt Ant International’s New AI-Powered Foreign-Exchange Tool

Citi, HSBC and Standard Chartered are among major global banks adopting Ant International’s upgraded artificial-intelligence platform for foreign-exchange forecasting and treasury management, highlighting how specialised AI is moving deeper into core banking operations.

Ant International has launched Falcon Time-Series Transformer Model 2.0, an upgraded financial forecasting model designed specifically for time-series data such as cash flows and foreign-exchange exposures.

The Singapore-based fintech company says six major global banks are working with the technology, including Citi, HSBC, Deutsche Bank, Standard Chartered and Barclays.

The expansion marks an important shift in financial AI. Banks are increasingly moving beyond general-purpose chatbots and employee productivity tools toward models designed for specific financial problems where improvements can be measured directly through lower hedging costs, stronger liquidity planning and more accurate forecasts.

Falcon TST 2.0 Targets Financial Forecasting

Ant International’s Falcon platform is built around time-series forecasting rather than general-purpose text generation.

Financial institutions constantly work with information that changes over time.

Examples include:

foreign-exchange demand,

cash balances,

transaction flows,

liquidity requirements,

and payment volumes.

Accurately forecasting those movements can have significant financial value.

Model Uses Financial Data to Predict Future Exposure

A multinational company may receive revenue in several currencies while paying suppliers in others.

Those flows rarely occur in perfectly predictable amounts.

Treasury teams therefore forecast how much currency they expect to receive or require.

Forecast Errors Increase Hedging Costs

Suppose a company expects to receive $100 million but actually receives only $70 million.

If it hedges the entire expected amount, it can become over-hedged.

The opposite problem occurs when actual exposure exceeds forecasts.

More accurate predictions allow treasury teams to hedge closer to real requirements.

That can reduce unnecessary transactions and working-capital requirements.

Ant Says Forecast Accuracy Exceeds 90%

Ant International says its Falcon technology can achieve more than 90% accuracy in hourly foreign-exchange demand forecasting in its own applications.

The company's platform also claims potential reductions of up to 60% in FX-related costs and working-capital requirements in selected use cases.

These figures are important because they frame AI adoption through measurable financial outcomes rather than abstract technological capability.

Falcon 2.0 Uses Mixture-of-Experts Architecture

The upgraded platform uses a Mixture-of-Experts architecture.

This approach divides specialised analytical work across different components of a model.

Specialisation Can Improve Efficiency

Instead of activating every part of a large model for every forecasting task, the system can direct particular data patterns toward relevant expert components.

This can improve computational efficiency while maintaining large overall model capacity.

Ant describes Falcon as a 2.5-billion-parameter model trained across roughly 300 billion time points.

Time-Series AI Differs From Large Language Models

Much of the public AI boom has focused on large language models.

Those systems are designed primarily to understand and generate language.

Financial forecasting presents a different problem.

Treasury Data Is Numerical and Sequential

A treasury model needs to understand how numbers evolve through time.

It may analyse:

seasonality,

payment cycles,

currency flows,

and recurring transaction patterns.

A model designed specifically for time-series information can potentially outperform a general-purpose language model on these tasks.

This is the commercial argument behind Falcon.

Citi Has Already Tested Falcon in FX Risk Management

Citi began piloting Ant International’s earlier Falcon TST technology in 2025.

The initial use case focused on airline customers selling tickets online across multiple currencies.

Airlines Have Complex Currency Exposure

A global carrier can receive ticket revenue in dozens of markets.

Customers pay in local currencies.

The airline may need to convert those revenues into another operating currency.

That creates continuous FX exposure.

Citi integrated Falcon with its Fixed FX Rates solution to improve forecasts of those currency flows.

Pilot Airline Reported Lower Hedging Costs

Citi and Ant International said the initial airline pilot reduced FX hedging costs by around 30%.

That result provided an early commercial demonstration of the technology.

The new Falcon 2.0 rollout builds on those earlier trials and expands the potential use of specialised AI across banking customers.

Citi Supports More Than 70 Currencies

Citi's Fixed FX Rates platform supports over 70 currencies and is used by businesses operating across travel, ecommerce and other international industries.

Adding predictive AI can make the service more dynamic.

Instead of hedging according to static assumptions, clients can potentially adjust exposure according to expected real-time transaction patterns.

Standard Chartered Integrates Falcon With SCALE

Standard Chartered has separately integrated Falcon with its Aggregated Liquidity Engine, known as SCALE.

The combination is designed to support real-time and 24/7 treasury and foreign-exchange management.

SCALE Provides Guaranteed FX Pricing

Businesses operating internationally may want predictable exchange rates even when transactions occur continuously.

Standard Chartered's system can provide FX liquidity and pricing across multiple currencies.

Falcon adds forecasting capability.

Together, the platforms can estimate future FX requirements and manage hedging accordingly.

Standard Chartered Reported More Than 90% Forecast Accuracy

The bank previously said integration with Falcon allowed it to forecast Ant International’s FX exposures with more than 90% accuracy.

This helped reduce hedging costs and improve cash-flow planning.

The partnership demonstrates how AI can be embedded directly inside established banking infrastructure rather than delivered as a separate analytics dashboard.

HSBC Expands Relationship With Ant International

HSBC is also among the banks working with Falcon 2.0.

The relationship builds on broader cooperation between HSBC and Ant International around digital treasury infrastructure.

The two companies have already worked together on tokenised deposit solutions and real-time cross-border treasury management.

AI and Blockchain Are Converging in Treasury

AI can forecast where cash will be needed.

Digital settlement infrastructure can then move that money.

Combining prediction with faster settlement could allow multinational companies to operate treasury functions more efficiently.

This represents a significant evolution from traditional corporate cash management.

Deutsche Bank and Barclays Also Participate

The adoption extends beyond Citi, HSBC and Standard Chartered.

Deutsche Bank and Barclays are also among the major institutions partnering with Ant International around the technology.

This provides the Falcon platform with access to several of the world's largest corporate and institutional banking networks.

For Ant International, bank distribution could become more valuable than selling directly to every individual corporate treasury department.

Banks Are Becoming Distribution Partners for AI

A corporate customer already trusts its bank to manage currency transactions.

Embedding AI within banking products can therefore reduce adoption friction.

Corporates Do Not Need Separate AI Infrastructure

Instead of buying a forecasting platform independently, a company can access AI-enhanced functionality through its existing bank.

This simplifies:

integration,

compliance,

and operational workflows.

Banks also gain an opportunity to differentiate their treasury products.

Corporate Treasury Is Becoming Real-Time

Traditional treasury management often involved daily or weekly forecasting.

Digital commerce has changed that environment.

Businesses can receive transactions continuously across countries and currencies.

Payments Now Operate Around the Clock

Ecommerce platforms, airlines and digital marketplaces can generate foreign-exchange exposures at any hour.

Treasury therefore increasingly needs:

continuous forecasting,

real-time liquidity,

and automated hedging.

AI can become valuable because manual forecasting struggles at that frequency.

Ecommerce Creates Particularly Complex FX Flows

Online merchants can sell simultaneously across dozens of markets.

Transaction volumes fluctuate according to:

promotions,

holidays,

consumer demand,

and regional events.

Predicting currency requirements using spreadsheets becomes increasingly difficult.

Time-series AI can analyse historical patterns at much greater scale.

Airlines Are Natural Early Users

Travel businesses experience similar complexity.

Ticket demand changes according to:

seasonality,

routes,

pricing,

and travel disruptions.

Revenue is received in numerous currencies.

Accurate forecasting can therefore directly reduce treasury costs.

This helps explain why Citi's first Falcon use case focused on aviation.

Digital Platforms Also Need Dynamic Currency Management

Marketplaces and payment companies process enormous volumes of cross-border transactions.

Even small improvements in FX execution can generate significant savings.

Scale Magnifies Minor Efficiency Gains

A 0.1% improvement sounds small.

Across tens of billions of dollars in transactions, it becomes economically meaningful.

This is why specialised financial AI can create substantial value without generating consumer-facing products.

AI Can Reduce Working Capital Needs

Companies often maintain extra cash to protect against uncertain future requirements.

Better forecasting reduces that uncertainty.

More Accurate Forecasts Free Cash

If a treasury department knows more precisely how much currency will be required tomorrow, it can keep smaller precautionary balances.

That cash can then be used elsewhere in the business.

Ant International says its platform can reduce working-capital requirements by as much as 60% in selected scenarios.

Liquidity Management Is Major Corporate Function

Large companies may maintain accounts at several banks across dozens of countries.

Cash can become fragmented.

One subsidiary may have excess money while another needs funding.

Treasury teams constantly attempt to optimise these balances.

AI can improve forecasts of where and when liquidity will be required.

AI Can Improve Hedging Timing

Foreign-exchange hedging is not simply about predicting currency prices.

Companies primarily hedge to manage exposure.

If expected cash flows are known more accurately, hedging decisions can be better matched to actual business activity.

This reduces the risk of both over-hedging and under-hedging.

Specialised AI May Gain Advantage in Banking

The Falcon rollout reflects a broader change in enterprise AI.

Early adoption focused heavily on general-purpose models.

Financial institutions increasingly want specialised systems.

Banking Requires Precision

A creative AI response can be acceptable in marketing.

A financial forecast has direct monetary consequences.

Banks therefore need models optimised for:

accuracy,

repeatability,

and risk controls.

This can create opportunities for smaller specialised models rather than relying exclusively on the largest general-purpose AI systems.

Financial AI Needs Explainability

Banks operate within highly regulated environments.

They cannot simply deploy black-box systems without understanding operational risks.

Human Oversight Remains Important

Treasury professionals need to know:

what data feeds the model,

how forecasts are produced,

and how errors are managed.

AI should therefore support decision-making rather than eliminate governance.

The stronger the financial impact, the more important model controls become.

Model Risk Management Will Expand

Banks already maintain extensive frameworks for credit and market-risk models.

AI introduces another layer.

Institutions need processes around:

validation,

monitoring,

bias,

and performance drift.

A forecasting model that works well today may become less accurate when market behaviour changes.

Continuous monitoring is therefore essential.

Extreme Market Events Remain Challenge

Historical patterns do not always predict financial shocks.

Events such as:

wars,

pandemics,

and sudden policy changes

can disrupt normal currency flows.

Models need mechanisms for recognising when historical assumptions are no longer reliable.

Human treasury expertise remains particularly important during such periods.

FX Forecasting Is Not Currency Speculation

There is an important distinction between forecasting corporate currency exposure and predicting whether the dollar will rise or fall.

Falcon is primarily designed to estimate expected business-related currency requirements.

That allows companies to manage hedging more efficiently.

The goal is generally reducing risk rather than making speculative bets on exchange rates.

Ant International Processes Huge Cross-Border Volumes

Ant International operates a large global payments and digital-commerce network.

The company supported more than 2 billion cross-border transactions in its core emerging markets during 2025.

This gives it access to extensive real-world time-series information.

Payments Data Can Improve Financial Models

High transaction volumes create large training datasets.

That can help models identify:

seasonality,

regional patterns,

and payment behaviour.

Data scale therefore becomes an important competitive advantage.

Ant International Is Expanding Beyond Payments

The company has traditionally been associated with cross-border payments and merchant technology.

Falcon demonstrates a broader strategy.

Ant International increasingly provides financial infrastructure involving:

AI,

treasury management,

and blockchain-based settlement.

This moves the company deeper into technology services for banks and multinational corporations.

Recent Funding Supports Expansion

Ant International recently raised approximately $1.2 billion in fresh equity financing as it seeks to expand internationally.

The adoption of Falcon by major global banks strengthens that expansion narrative.

Rather than competing directly with banks, Ant increasingly positions itself as a technology partner supplying infrastructure banks can use with their own customers.

Partnership Model Could Accelerate Scale

Selling enterprise software one customer at a time can be slow.

Partnering with major banks provides access to thousands of multinational clients.

If Falcon becomes embedded in treasury products, Ant International can potentially scale through banking networks.

This makes institutional partnerships strategically valuable.

Banks Gain Competitive Differentiation

Corporate banking products can sometimes appear similar across institutions.

AI creates another area for differentiation.

A bank offering more accurate FX forecasts can potentially help clients reduce costs.

That can strengthen customer relationships.

Treasury Technology Can Influence Bank Selection

Large corporations evaluate banks based on more than lending rates.

They also consider:

payments,

FX execution,

liquidity tools,

and technology.

Advanced AI capabilities can therefore become a competitive advantage in transaction banking.

Transaction Banking Is Strategically Valuable

Corporate treasury relationships can be highly profitable for banks.

A company using one bank for global payments may also purchase:

foreign exchange,

trade finance,

and lending.

Treasury technology can therefore deepen broader corporate relationships.

This explains why major institutions continue investing heavily in payments infrastructure.

AI Could Automate More Treasury Decisions

Today's systems primarily support forecasting.

Future platforms could potentially automate greater portions of treasury workflows.

For example, a system could:

predict exposure,

select hedging instruments,

and execute predefined transactions.

That would move corporate finance closer to autonomous operations.

Full Automation Will Require Strong Controls

Large currency transactions carry financial risk.

Businesses will likely retain approval thresholds and human oversight for significant decisions.

AI may handle routine exposures automatically while escalating unusual situations.

A hybrid approach is more realistic than eliminating treasury teams entirely.

Treasury Professionals Could Shift Toward Strategy

As AI automates repetitive forecasting, employees can focus more on:

capital allocation,

risk strategy,

and financing.

This mirrors the broader impact of AI across professional services.

Technology changes the type of work rather than simply removing every role.

Banking AI Is Moving Into Core Operations

Financial institutions initially used AI heavily for:

fraud detection,

customer service,

and document processing.

Foreign-exchange forecasting moves AI directly into revenue-generating markets businesses.

That is strategically significant.

The technology is no longer operating only around the edges of banking.

FX Market Is Enormous

Foreign exchange is one of the world's largest financial markets.

Trillions of dollars change hands every day.

Even modest improvements in execution and hedging can therefore create enormous aggregate savings.

This makes the sector an attractive target for specialised AI.

India Could Benefit From AI Treasury Adoption

Indian multinational companies increasingly operate globally.

IT services firms, pharmaceutical companies, manufacturers and exporters all manage foreign-currency exposure.

More sophisticated treasury tools could help these businesses reduce volatility and improve cash management.

Rupee Volatility Creates Real Business Risk

An exporter may earn dollars but pay costs in rupees.

An importer faces the opposite exposure.

Changes in exchange rates can therefore affect margins substantially.

AI-enhanced forecasting cannot eliminate currency volatility, but it can help companies manage the timing and size of hedges more accurately.

Indian Banks May Eventually Adopt Similar Tools

Global institutions such as Citi, HSBC and Standard Chartered already operate extensively in India.

Their adoption of AI-driven treasury technology could influence local corporate banking practices.

Domestic banks may also increase investment in specialised forecasting models as customers demand more advanced services.

Fintech Competition Could Accelerate Banking Innovation

Ant International's strategy demonstrates how fintech companies can supply technology without replacing banks.

This partnership model can accelerate innovation.

Banks retain:

regulatory relationships,

customer trust,

and balance sheets.

Fintech companies contribute specialised technology.

Combining the two can be more effective than direct competition.

Data Privacy Remains Critical

Treasury forecasting uses sensitive corporate information.

Banks and technology providers need strict controls around:

transaction data,

cash-flow information,

and customer confidentiality.

Corporate clients will need confidence that proprietary financial information remains protected.

Cybersecurity Is Equally Important

AI systems integrated into treasury operations become valuable targets for cyber attackers.

Manipulating financial forecasts or payment instructions could create substantial losses.

Security therefore needs to be built into the system architecture.

AI accuracy has little value without secure deployment.

Banks Need Audit Trails

Automated systems must record how decisions were generated.

Auditability becomes particularly important when transactions affect regulatory reporting or financial statements.

Institutions need clear records showing:

what data was used,

which forecast was produced,

and what action followed.

This is another reason enterprise financial AI differs significantly from consumer chatbots.

Cost Savings Could Drive Adoption Faster Than AI Hype

The most commercially important aspect of Falcon may be its measurable economics.

Businesses do not necessarily need to understand the underlying model architecture.

They need to know whether it saves money.

Finance Departments Respond to ROI

If an AI system consistently reduces hedging costs and frees working capital, adoption becomes easier to justify.

This can create a faster enterprise sales cycle than technologies whose benefits are difficult to quantify.

Specialised Models Could Become Major Enterprise AI Category

The financial industry may be an early example of a broader trend.

Companies may increasingly prefer narrow models designed for specific tasks.

Examples could include:

demand forecasting,

credit risk,

and industrial maintenance.

These models can be smaller than general-purpose systems but more useful within defined workflows.

General-Purpose AI Will Still Play Role

Specialised models do not eliminate broader AI platforms.

Large language models remain useful for:

research,

communication,

and workflow orchestration.

The most effective enterprise architecture may combine different types of models.

A language model can interact with users.

A time-series model can perform numerical forecasting underneath.

This layered approach could become increasingly common.

Falcon Could Expand Beyond FX

Time-series forecasting applies to many financial problems.

Potential applications include:

liquidity,

treasury balances,

and transaction demand.

Ant International may therefore expand Falcon into additional banking workflows over time.

The current FX rollout provides a practical starting point because the financial benefit is relatively easy to measure.

Bank Adoption Provides Important Validation

Citi, HSBC and Standard Chartered are highly regulated global institutions.

Their willingness to integrate Ant International technology provides meaningful validation.

Enterprise financial buyers are generally cautious about deploying new models into core systems.

Successful implementation with major banks could make Falcon easier to sell to additional institutions.

Competition Will Intensify

Ant International will not have the market to itself.

Banks already develop proprietary AI internally.

Cloud companies also offer time-series forecasting tools.

Specialist fintech companies are building treasury platforms.

Falcon therefore needs to maintain:

accuracy,

security,

and integration advantages.

Conclusion

The adoption of Ant International’s Falcon TST 2.0 by Citi, HSBC, Standard Chartered and other major banks marks an important stage in the evolution of artificial intelligence inside global finance.

The platform is designed specifically for time-series forecasting, allowing banks and corporate customers to predict foreign-exchange exposures and liquidity requirements more accurately.

Ant International says the system can achieve more than 90% forecasting accuracy in selected use cases and potentially reduce FX-related costs by over 60%.

Previous deployments already provide practical evidence. Citi used Falcon in an airline FX-risk pilot that reduced hedging costs, while Standard Chartered integrated the model into its real-time SCALE treasury infrastructure.

The broader significance lies in where AI is being deployed.

Banks are moving beyond employee copilots and customer-service chatbots toward specialised models embedded directly into treasury, payments and foreign-exchange operations.

For Ant International, partnerships with major financial institutions provide a powerful distribution channel for its technology.

For banks, the opportunity is equally clear: better forecasting can reduce customer costs, deepen corporate relationships and differentiate increasingly digital transaction-banking businesses.

If specialised AI models continue producing measurable savings, treasury management could become one of the first areas where artificial intelligence moves decisively from experimental banking technology into core financial infrastructure.