As global interest rate volatility heightens, the importance of effective and accurate cash forecasting has increased, prompting chief financial officers and treasury teams to seek out new tools to increase their confidence in cash projections.
Treasurers are increasingly seeking automation coupled with real-time data and predictive analytics not only to accelerate cash decision making but also to transform how they are building and measuring their forecasts. The allure of artificial intelligence (AI) is its power to both automate and transform. It is, for every CFO and treasurer that we speak with, an opportunity they believe they must embrace to avoid being left behind.
Fortunately, the adoption of AI is gaining significant traction locally and worldwide. In the United Arab Emirates (UAE) alone, the AI market was estimated at $3.47bn in 2023, and is expected to reach $5.22bn by 2024. This growth underscores the increasing recognition of AI's potential to drive efficiency, accuracy and strategic decision-making across the organisation, including finance.
Cash-flow forecasting is crucial for the office of the CFO, but traditional methods have often fallen short. While treasury workstations and ERP platforms have offered cash forecasting capabilities for decades, the majority of treasury teams continued to rely on manual work and complex spreadsheets, limiting the ability to adapt forecasting models to new data and evolving forecast scenarios – such as unpredictable inflation or changing interest rates. The top five reasons why treasury teams have been unable to rely on treasury technology to deliver accurate forecasting have historically been:
AI is revolutionising cash forecasting by providing data-driven insights built into tools that can handle today’s forecasting complexities. By leveraging AI, organisations can better navigate economic uncertainties and make informed decisions that optimise cash and liquidity alongside improved bottom-line impacts.
The path to AI begins with data. There is no AI strategy without a data strategy. CEOs and their executive teams recognise the need for data, with many organisations investing in data scientists and data strategies to unlock more data and organise ways to feed AI models. Within our own customer base alone, we observed a 66% increase in bank transactions – a key input for AI-driven cash forecasting models – between 2023 and 2024.
A well-defined data strategy specifies how a company collects, stores, manages and analyses its data, which is essential for the successful integration of AI into financial operations.
A treasury team can start by asking these questions:
The most challenging part of implementing this basic AI for forecasting is the training of the AI models
AI can aid cash forecasts in three ways:
The most obvious benefit of AI is to improve the accuracy of forecasting. AI identifies patterns in historical forecast data that can be used to better predict future forecast movements. This is the most basic use case, which utilises machine learning, trained in prior forecast datasets – most particularly, how forecasts changed over their lifetime. For example, feeding the AI model with forecast data at six months, three months, one month, one week and, finally, how it reconciled with bank actuals during the prior-day forecast variance offers tremendous insight into the future behaviour of forecast cash flows. Patterns emerge across thousands of transactions that AI can more quickly identify – and then apply to the office or the CFO’s 13- and 26-week forecasts.
The most challenging part of implementing this basic AI for forecasting is the training of the AI models. In this case, it is recommended to manage multiple cash forecasts, with AI models pitted against more traditional forecasting data, and the two versions compared against each other to confirm where AI was – or was not – more accurate.
Treasury technology that supports multiple, parallel forecasts will simplify the time management involved, avoiding what could be a very difficult setup and review.
Generative AI (GenAI) continues to evolve in way that will see it play a role in improving forecasting – notably, to streamline the process of training and evaluating AI models. The conversational capabilities of GenAI tools (such as ChatGPT) lend themselves to simply asking your cash forecast where AI is more effective.
Combining with additional AI tools embedded in software packages (such as Microsoft Copilot within Excel) means that the treasury team responsible for forecasting can offload the execution, evaluation and presentation of cash forecast results to AI. This frees up time for treasury teams to focus more on actionability of the now improved forecasts, identifying and actioning more optimised liquidity decisions.
As market volatility becomes the norm and finance teams are asked to understand and apply significantly greater quantities of data, traditional ‘non AI’ forecasting methods mean that treasury teams often find themselves operating one step behind. Staying ahead of the curve demands a more efficient, accurate, and dynamic approach to cash forecasting – one that AI is uniquely positioned to deliver.
Bob Stark is the global head of market strategy at Kyriba
The views in this article are those of the author and do not necessarily represent the views of the Association of Corporate Treasurers
