How AI is changing treasury management (and what comes next)

AI in treasury management and cashflow forecasting

Last updated: 18 September 2026

The AI revolution continues to gather pace, transforming entire industries and roles almost overnight.

Treasury management has involved bringing together large quantities of information, such as bank balances, cashflow forecasts, currency exposures, payments, invoices, debt and market data, before using it to make decisions. The difficulty was always the sheer amount of data required to make sensible decisions, data that often sits across different systems, spreadsheets, bank accounts and business units.

AI is increasingly capable of doing some of that hard work.

While we probably won’t see an autonomous treasury department anytime soon, AI has the potential to give finance teams better visibility, automate repetitive work and identify risks earlier. And we are still relatively early in that transition.

How is AI being used in treasury management?

AI can be used to carry out several treasury functions. Machine learning can identify patterns in large datasets and improve forecasts. Generative AI can interrogate information and communicate findings in natural language. Increasingly, AI agents are capable of carrying out complex tasks within predetermined rules.

PwC’s 2025 Global Treasury Survey found that 74% of respondents were either expanding or actively using AI, with machine learning and predictive analytics receiving particular attention. However, only 26% described their AI capabilities as moderately or very mature.

AI may already be changing treasury, but it is also very new. There is a considerable difference between experimenting with a tool and building it into the way a finance function operates.

Still, many companies are using it to manage specific treasury tasks.

1. Cashflow forecasting

Cashflow forecasting is an obvious application for AI because businesses already possess enormous amounts of data that could potentially tell them something about their future cash position.

Traditional forecasting might use historical figures alongside assumptions supplied by different parts of the business. AI models can potentially analyse much larger datasets, identify recurring patterns and continually incorporate new information.

That could include customer payment behaviour, seasonal demand, supplier payments, payroll, outstanding invoices and historical forecast errors.

Rather than replacing the cashflow forecast, AI can therefore make it more dynamic.

Traditional forecasts inevitably become less useful as the assumptions behind them age. A system capable of continuously incorporating new information could help finance teams identify a potential cash shortfall or surplus considerably earlier.

Recent research from EY illustrates where the technology could be heading. Its September 2026 study argued that agentic AI-enabled treasury models could potentially deliver forecast accuracy of up to 90% across 30-, 60- and 90-day liquidity horizons, although results will naturally depend on the organisation, data and implementation.

2. Finding important information

One of AI’s less glamorous applications may ultimately be among its most useful.

Treasury teams can spend significant amounts of time gathering, cleaning and reconciling information before analysis can even begin. It is often a huge administrative burden for growing companies.

Instead of manually reviewing thousands of transactions, for example, technology can categorise payments and flag unusual movements for investigation. It can help reconcile information from different sources or identify anomalies that might otherwise require somebody to work through a spreadsheet.

The objective is not simply to complete the same work faster. It is to give treasury professionals more time to investigate exceptions, model scenarios and make decisions. For businesses with relatively small finance teams, the productivity implications could be significant.

3. Managing currency risk

A company with international revenues or costs may have exposures spread across multiple currencies, subsidiaries, invoices and future commitments. The immediate challenge is therefore not necessarily deciding what to do about an exposure. It is establishing exactly what the exposure is.

Automation can help consolidate that information. AI can go further by identifying patterns, testing scenarios and highlighting deviations from expected cashflows.

Imagine, for example, a UK business expecting to make a series of euro payments over the next six months. Rather than periodically updating a spreadsheet, an integrated system could continually compare expected payments, existing hedges and changes to the underlying forecast.

If an exposure changed materially, the finance team could be alerted. That creates the possibility of moving from periodic currency risk reviews towards much more continuous monitoring.

The distinction between information and judgement, however, remains crucial. An AI system might identify an exposure or model the potential financial effect of a currency movement. Deciding how much risk a business should accept (as well as how that fits its wider commercial objectives) still requires human expertise.

4. Scenario modelling

What happens if sterling falls 10%? Or if customers begin paying two weeks later? What about when borrowing costs increase, or when a major supplier tweaks its payment terms?

Finance teams have always been able to model scenarios like these. The difference AI offers is speed and scale.

Instead of manually constructing a handful of scenarios, increasingly sophisticated systems can analyse numerous variables simultaneously and show how changes could flow through liquidity, working capital and currency exposure.

This could become particularly useful during periods of disruption.

When market conditions change rapidly, a forecast produced several weeks ago can quickly become outdated. AI-supported treasury could allow businesses to continuously reassess their assumptions as new information arrives.

From AI assistant to AI agent

The next stage of this development might be even more significant.

Most current AI applications assist people with tasks. Agentic AI is designed to complete sequences of tasks itself within defined parameters. For treasury, that creates some interesting possibilities.

Future AI-native treasury platforms could continuously monitor balances and exposures and potentially perform actions such as liquidity sweeps within predetermined governance controls.

Consider a multinational business with cash distributed across multiple accounts. An AI system could theoretically monitor those positions, forecast near-term requirements and recommend how surplus liquidity should be allocated. More advanced systems could eventually carry out approved actions automatically, escalating exceptions to treasury staff.

Treasury technology has historically helped humans understand what has happened. The next generation could increasingly identify what is likely to happen and initiate an appropriate response.

Better data > better AI

Of course, AI cannot magically repair poor financial information.

If a company’s cash data is fragmented, forecasts are inconsistently maintained or currency exposures are not properly captured, adding AI may simply produce sophisticated analysis of unreliable inputs. PwC’s treasury research found poor data quality was cited by 76% of respondents as a challenge to forecasting.

This means businesses considering AI should first ask whether they have reliable and appropriately governed data that is readily available to relevant stakeholders.

The foundations of good treasury management (accurate information, clear policies, defined responsibilities and appropriate controls) become more important rather than less important when automation is introduced.

Will AI replace the corporate treasurer?

It seems more likely to change their role.

AI is particularly effective at processing large amounts of information, identifying patterns and completing repetitive tasks. Humans remain important where decisions require commercial context, accountability and judgement.

A currency exposure provides a useful example. Two businesses might have an identical £5mn euro requirement but make entirely different decisions about managing it. One may operate on extremely thin margins and prioritise certainty. Another may have substantial cash reserves and greater tolerance for short-term volatility.

The numbers alone cannot define the appropriate approach.

AI can help a treasury team understand those numbers more quickly and explore the consequences of different scenarios. It cannot determine a company’s objectives or risk appetite on its behalf.

The more interesting future, therefore, is not humans versus machines. It is treasury professionals using better technology to spend less time assembling information and more time deciding what to do with it.

What does the future of AI in treasury look like?

Over the next few years, the distinction between “AI treasury technology” and ordinary treasury technology may begin to disappear.

Cashflow platforms, treasury management systems, banks and financial service providers are increasingly incorporating machine learning and generative AI into their products. Eventually, AI may simply become part of the furniture.

That could move treasury towards a more continuous model. Cash positions could update automatically. Forecasts could change as new transactions arrive. Currency exposures could be reconciled against hedges. Unusual payments could trigger alerts. Scenario models could continually test the resilience of the company’s liquidity position.

The treasury professional would increasingly sit above that system: setting parameters, interrogating its conclusions and making the decisions that require human judgement.

That is potentially the most important change AI will bring to treasury management.

For decades, one of the biggest challenges facing finance teams has been turning a huge quantity of financial information into a relatively small number of good decisions. AI cannot remove uncertainty from those decisions, but it may dramatically improve the information available when they are made.

Smarter treasury starts with better visibility

Smart Currency Business helps UK companies understand and manage their currency exposure as part of their wider treasury requirements. Combining specialist support with treasury technology, we help finance teams gain greater visibility over their exposures, cashflows and currency risk.

Speak to our team or register with Smart Currency to discuss how a structured approach to treasury management could help your business.