Corporate treasury is responsible for cash, liquidity, foreign exchange, commodity exposure, investment, and compliance. Yet some departments still manage these decisions through spreadsheets that sit between trading platforms, banks, and enterprise accounting systems.
That manual handoff is the first problem an artificial-intelligence programme must solve. If employees copy a trade from one platform into a spreadsheet and later post an accounting entry into an enterprise resource planning system, the data may be delayed, incomplete, or inconsistent before a model ever sees it.
Ashish Kumar of Infosys and CM Grover of IBS FinTech argue that treasury AI depends on digitised, automated data. The technology cannot create financial resilience from a workflow whose current state is not reliably visible.
Real-Time Decisions Require Connected Records
Treasury teams operate in markets that can change faster than a reporting cycle. Currency movements affect imports and exports, commodity prices alter cost exposure, and surplus cash needs to be invested without weakening liquidity.
A connected treasury-management system can exchange information directly with trading platforms, banks, and the enterprise resource planning environment. This reduces re-entry and gives executives a current view of positions, cash, risk, and compliance.
IBS FinTech built its backend on Oracle databases and integrates with Oracle Cloud, NetSuite, and Fusion. The specific products are less important than the architectural principle: each transaction should move through authorised systems without requiring a person to recreate it.
AI can then forecast liquidity, identify unusual exposure, or prioritise action using a consistent data trail. Without that foundation, a sophisticated model may simply automate a spreadsheet error.
Automation Creates the Basis for Intelligence
Leaders sometimes begin with an AI use case when they should begin with a workflow audit. They need to map where a trade originates, which fields are retyped, how reconciliations occur, when accounting is updated, and which system owns the final record.
The first returns may come from integration rather than prediction. Eliminating manual transfer can improve timeliness, reduce errors, and create an audit trail even before a model makes a recommendation.
Once the pipeline is dependable, AI can support scenario analysis across foreign exchange, commodities, and liquidity. It may identify how combined movements affect cash needs or show where a risk limit could be breached.
The model should not silently execute high-impact treasury actions. Teams can begin with alerts and recommended hedges, compare those outputs with existing decisions, and expand authority only after performance is understood.
Useful measures include forecast accuracy, manual entries removed, reconciliation breaks, response time to exposure, compliance exceptions, and the amount of verification work still performed outside the system.
Financial Resilience Depends on Control
Grover expects geopolitical and economic volatility to increase across commodities, equities, and foreign exchange. Faster information can help companies respond, but speed without governance can also move an error more quickly into a trade or ledger.
Connected systems need role-based access, approval thresholds, traceable data lineage, and recovery procedures. Treasury leaders should know which information shaped a forecast, when it was updated, and whether an unusual market condition falls outside the model’s experience.
Integration also concentrates operational dependence. If the treasury platform becomes unavailable, the company needs a tested fallback that preserves both decision-making and accounting integrity.
The promise of treasury AI is not a magical prediction of every market move. It is a more complete and timely view of cash and risk, followed by better-supported decisions.
That promise begins with unglamorous engineering. Direct connections must replace spreadsheet handoffs, definitions must be reconciled, and authoritative records must stay clear. Only then can AI improve treasury rather than decorate it. In finance, the path to intelligence often starts by ensuring that the same number no longer has to be typed three times.
