In today’s fast-paced, macroeconomic climate, treasurers are turning to artificial intelligence (AI) and machine learning (ML) tools in payment transactions and trade finance to adapt and react to market volatility. These tools also enable teams to prioritize time and resources more efficiently.
Increasingly, AI agents are underpinning the transformation towards real-time treasury. They reduce manual workloads by digitalizing treasury processes, and automate repetitive tasks, such as reconciliation and cash tagging. However, for treasurers able to lay the correct high-quality data foundation, AI and ML offer ROI gains beyond automation. When deployed correctly, these technologies have significant implications for improving fraud detection and providing near-instant analytical insights.
Automation transformation
With modern treasurers seeking strategies to free up valuable team time for more analytical, strategic work, AI and ML are being rolled out across the treasury function to simplify repetitive manual tasks. Generative AI (GenAI) has also proved highly useful, allowing treasurers to analyze databases and summarize results quickly, or read and summarize policies.
However, the application of AI to the treasury function runs much deeper than copilot tools and GenAI. Agentic AI automates foundational treasury processes across the treasury management system (TMS), and other systems that interact with the TMS, such as an enterprise resource planning (ERP) system. Using agentic AI for cash tagging and fraud detection processes removes the need for pre-defined rules, and requires even less manual intervention. These agents can be trained on a company’s proprietary financial data, allowing them to learn the specifics of organizational data and business needs.
By deploying AI- and ML-based automation tools, companies can automate financial processes across the business, and move away from manual, labor-intensive spreadsheets and bank portals.
AI day-to-day
The day-to-day efficiency gains of AI and ML should not be understated, even in these early years of uptake. With automation technology, daily bank reconciliation is no longer a tedious manual task requiring complex resolutions of transaction discrepancies. Bank fee codes are insufficient for the detailed categorization required, making it difficult to match reported transactions with expected transactions. In addition, unclear transaction descriptions and varying data formats hinder matching accuracy and lead to lengthy investigations.
Likewise, dealing with vast amounts of financial data from multiple sources is time-consuming and overwhelming, especially with the continuous flow of new data and timing of cash movements. Companies often struggle with pulling data from multiple systems and relying on inconsistent and inaccurate forecast submissions from different business units.
However, AI and ML treasury solutions equip treasury teams to automate exception handling, prioritize solutions, and improve the efficiency and accuracy of reconciliations and forecasts. Importantly, automation minimizes human error and reduces the need for costly and time-consuming manual corrections.
From data to forecasts
In addition to the efficiency benefits, AI and ML are proving transformational to analytics and predictive work as well. Pivoting and capitalizing on market volatility is one of the most vital aspects of a treasurer’s role. To do this successfully, Treasurers must make more frequent forecasts and with a shortened forecast window. For organizations facing skills or staff shortages, it’s a challenge to undertake more cash forecasts within tighter timeframes, with access to fewer resources.
AI and ML algorithms have proven transformational, rapidly analyzing and predicting patterns from large volumes of cash transaction data from different sources – such as accounts payable, accounts receivable, or bank account balances. Neural networks can increase forecasting accuracy, while repetitive tasks – such as data processing and analysis – are carried out faster and more effectively, streamlining the cash forecasting process. Indeed, ION Treasury data shows that cash forecasts can be done 3,000 times faster using ML techniques.
Using AI and ML to provide an accurate view of global cash positions across multiple bank accounts and subsidiaries in real time, treasurers can make better informed decisions related to cash flow management. AI-powered transaction tagging models can classify transactions correctly, and to detect payment outliers through deviations from learned behaviors and patterns. Treasurers can then compare or back-test generated forecasts against actual cash flows to see where the patterns inferred by AI and ML can enhance their decision-making.
Data quality
However, the efficacy of these agentic models is inherently tied to the organizational quality of the data they are trained with and operate on. Treasurers must also ensure that data elements and information-gathering systems are appropriately configured to support the necessary decision-making processes. This is the case with cash flow types, where ML can also help by automatically tagging transactions without the need for pre-defined rules.
Robust cybersecurity guardrails must accompany the integration of AI tools. By definition, treasury holds sensitive information, which poses the risk of a data leakage if any available Large Language Model (LLM) is used. Both the TMS vendor and the customer need to be aware of this situation, and take precautions to protect sensitive information, and intellectual property (specifically for the vendor).
GenAI and agentic AI have begun to revolutionize the treasury function, driving massive efficiency gains both in day-to-day tasks and through automating entire foundational processes. However, for ML and AI to be effective, it is crucial for a treasury function to own well-structured and relevant data that is codified to the enterprise’s needs.
