Artificial intelligence (AI) is touching every aspect of our lives and treasury is no exception. Many treasurers across the globe are exploring the potential use cases, as well as the challenges, that the technology presents. “Treasurers can’t ignore AI,” says Royston Da Costa, Assistant Treasurer at US distribution company Ferguson, who is also a Senior Lecturer on AI at the Centre for Finance, Technology and Entrepreneurship, in partnership with the Association of Corporate Treasurers.
He believes that treasury departments need to embrace AI as part of the current wave of technology change. “In treasury, we are at a pivotal point in the way in which technology is playing an increasingly huge part in our day-to-day operations,” he adds.
According to Strategic Treasurer and Cash Management Leadership Institute’s 2024 Generative AI (GenAI) in Treasury and Finance Survey report, treasurers are “unequivocal” in their support for AI and GenAI.
The research found that while only 25% of corporate treasury respondents considered themselves to be ‘very familiar’ with AI, the vast majority (83%) of those currently experimenting with the technology expected to expand their use.
Cash flow forecasting (CFF), a perennial problem for treasurers, is the top challenge that those polled think AI will address in the near term. Interestingly, more than six out of ten (62%) treasurers were confident that they would roll out AI technology to support CFF needs within two years, while 35% thought they could do it within one year. The survey found that one in 11 are already piloting AI technology for CFF.
Optimising treasury
Da Costa agrees that CFF is an area ripe for AI. “AI has the capability to analyse vast amounts of data. It not only looks at historical data, but can also forecast complex scenarios,” he says. “However, not every corporate will jump into using AI for CFF without having a better knowledge of how to use and apply it effectively.”
According to Elena Strbac, Global Head, Data Science and Innovation, Corporate and Institutional Banking at Standard Chartered Bank, AI’s potential to predict future cash flows will allow treasurers to better anticipate liquidity needs of the business, avoid cash shortfalls, minimise borrowing costs and optimise cash reserves. “The increasing use of application programming interfaces (APIs) among corporate treasuries is an important enabler of AI-powered cash and liquidity models, as real-time data such as account balances and alerts will be key inputs to AI predictions,” she adds.
Expanding on Da Costa’s point, Strbac says explains AI algorithms can ingest multiple internal and external data sources including balances, alerts, historical cash flows, accounts receivable/accounts payable, macroeconomic indicators, and foreign exchange (FX) volatilities.
“Algorithms can then be trained to predict cash flows, trends and anomalies, or to predict reconciliation discrepancies between payments and invoices in the enterprise resource planning system – ultimately allowing treasurers to better optimise their resources,” she says. This is important in the face of high market volatility and uncertainty, according to Nils Bothe, Partner, Financial Services, Finance and Treasury Management at KPMG.
“Treasurers are more mindful of their liquidity to cope in times of crisis. Being able to accurately predict cash flows is critical for effective financial management to make informed business decisions and for eventual long-term success,” he says. “With AI and machine learning, it becomes easier to create automated and accurate forecasts to minimise risk and optimise decision making.”
In a centralised cash management structure like Ferguson’s, AI could prove useful in applying advanced analytics in dynamic liquidity planning and working capital optimisation, according to Da Costa. In addition to CFF, Strbac thinks that AI could be leveraged for stress testing, scenario and sensitivity analysis, to predict the impacts of external market factors and internal financial strategies on the overall liquidity.
Bothe agrees, saying that treasurers are ideally looking for “stress testing at a button”. He reports that key treasury management system (TMS) providers are integrating AI into their cash management and liquidity planning modules, making use of modelling options such as autoregressive integrated moving average, additive regression and neural network regression. “Eventually these tools will evolve to cover complex relationships that may be overlooked by humans. There is value in human expertise, but it can now be combined with advanced AI tools to optimise predictions and prepare for unforeseen events,” he adds.
Adoption maturity
As indicated by the survey, most treasuries are predominantly at an experimental stage in their AI journeys. According to Strbac, this means they are focusing on developing frameworks that will enable them to measure the impact and success of AI initiatives, identifying the highest priority use cases and building technology stacks that will meet their future AI needs. “More advanced corporate treasuries are already embracing AI, particularly for anomaly and fraud detection,” she says. “They are also making advancements with process automation, such as payment reconciliation and payment processing. Traditional rules-based approaches to process automation can still be effective, but combining these with AI leads to more scalable solutions.”
Strbac uses the example of checking the quality, completeness and accuracy of payments messages against complex policies and standards. “This may involve a number of deterministic checks, but more complex rules or common errors can be learnt by an AI algorithm,” she says.
“AI can then produce a probabilistic alert for a potentially more complex issue to be investigated, providing a smarter layer on top of the rules-based approach that can lead to better data-driven decision making.” Interestingly, Ferguson is looking at implementing AI for streamlining a process involving requesting approvers to approve a payment within its TMS solution, reports Da Costa.
KPMG is talking to clients about using specific AI methodologies to identify fraudulent payment patterns that would otherwise be overlooked by humans, according to Bothe. He also points to other use cases, such as transaction tagging, which is matching cash transaction data with bank statement items, and algorithmic trading, which uses AI to autonomously identify and execute trades, manage the risks and order flow. “[The latter] also improves liquidity management by dynamically optimising size, duration and order size based on current market conditions,” he adds.
Deployment challenges
In addition to opportunities derived from deploying AI, there are also risks and challenges that need to be factored in. According to KPMG’s observations, treasurers remain hesitant to move from the exploration phase to rolling out AI because of data sensitivity concerns, as well as identifying the most value use cases. “The treasury function is a central point of contact for financial information and key figures, and the department routinely collects and processes highly sensitive data. Therefore, treasury needs to pay careful attention to data protection requirements,” explains Bothe.
Any deployment of cloud solutions, which are frequently used due to the large volumes of data involved in AI technologies, should be treated with the respective caution, he warns. “It is important to check whether sensitive personal data is stored, where the server is geographically located, whether the data storage meets the company’s specific protection requirements, etc,” he adds. Unsurprisingly, data protection concerns are paramount when considering ChatGPT and similar publicly accessible AI-powered large language models.
In addition, new ethical demands are placed on treasury due to the use of AI. “Any deployment should be reviewed to determine the extent to which unintended bias or discrimination could occur,” Bothe says. He points to the use of AI-based models to predict credit risk, where the models can make predictions based on historical data and other factors such as demographic characteristics and credit behaviour patterns. “These also give rise to new legal issues in terms of liability for AI-based decisions in treasury,” he cautions.
Picking up Bothe’s point about data bias, Da Costa adds that it is important for treasurers to remember that AI isn’t perfect. “It’s flawed by the fact that the data is not always accurate nor diverse. If you look at forecasts that are always skewed to a demographic that had been inputting that data, then you’re not going to get the best possible outcome from that data,” he says. “However, that’s slowly changing.”
Strbac identifies other key challenges for treasurers to navigate, such as recruiting talent with skills in machine learning, reinforcement learning and GenAI, in locations that are aligned to business operations and footprint.
From her perspective, working with the organisation’s technical architects to ensure scalable tech design for the organisation’s AI journey, and articulating the roles and responsibilities in the end-to-end AI lifecycle, in particular between the business, functions, technology and chief data office teams, are also important. She believes that banks can help corporate clients navigate many of these challenges. “Banks can share their knowledge in leveraging AI technologies in finance in a form of advisory service, bringing their in-house AI experts who can provide hands-on guidance. These same teams can work with the corporate treasuries to co-create new AI solutions designed and customised for specific treasury needs,” she says.
A potential area for future co-creation could be creating tailored, AI-powered process-check engines between the bank and the client, suggests Strbac. “These engines could identify settlement instructions, payment instructions or amendments; check for completeness against pre-defined criteria; and suggest corrections, before the process fails and requires manual intervention,” she explains. “This can lead to improved client experience and an advancing banking proposition that supports corporate treasurers in their move to real-time and just-in-time treasury.”
Long-term perspective
One of the biggest challenges is staff’s concerns that AI will make them redundant – a particular worry in small treasury teams. However, Da Costa is sceptical this will happen. “AI is not going to make anyone redundant – instead it will enhance our roles in treasury,” he says. “It will definitely eliminate the more mundane routine processes, but our roles will evolve to become more analytical.” While Bothe believes that perhaps certain job profiles characterised by highly repetitive tasks could be overtaken by an AI system to a certain extent, he believes that there still needs to be the “human factor” for interpreting and making well-informed decisions, and to turn to more strategic tasks. “The interesting challenge in the future will be how to ensure that this level of expertise is always available in a company when the repetitive tasks are being done by technology,” he says.
The biggest impacts of AI are expected to be in increasing productivity and unlocking operational efficiencies, according to Strbac. “For treasuries, AI has the potential to drive a more personalised and efficient experience with their banks, for example with digital banking platforms being tailored to facilitate the predicted next best action for treasurers to take, and automatic generation of API implementation code enabling faster connectivity,” she says. “In future we may also see banks creating finance-specific AI products, just as tech companies make their AI products accessible to other institutions to use and customise to their needs. This could include AI powered models for predicting early alert default propensity, FX market liquidity or currency volatility,” she adds.
Treasurers should be thinking about what they want to achieve in future and how technology can support their journey, advises Da Costa. “Ferguson, for example, has a fairly high level of automation, but we began preparing ourselves for the future – which is closer than you think,” he says. “The speed in which technology is evolving means that creating a ten-year plan is impossible, but we can think about the infrastructure we will need to future-proof treasury – and AI will form a big part of that.”
