Treasury Digitization - Market Perspectives

9 Treasury Digitization: Market Perspectives Figure 6: Applications for Which ML/AI is Being Explored or Pursued Other Implementing more Sophisticated Risk Management Strategies Supporting Decisions for Debt/Investment Management/Yield Optimization Identifying and Quantifying Risk Exposures with Greater Speed and/or Accuracy Detecting Outliers or Fraud within Payments/Trading Improving the Timeliness and Accuracy of Cash Forecasts Automating Cash Application and Collections Reconciliation 8% 12% 13% 14% 23% 35% 48% Number of Respondents: 214 Supporting More Agile Decision Making As the evolution in payments propels treasury towards an instant 24/7 environment, we anticipate increased demand for predictive or even prescriptive analytics to support and enhance decision-making processes. ML and AI technologies have the potential to drive a smarter treasury operation and are of interest for many treasurers, as shown in Figure 6 . Almost half the survey respondents are considering ML/AI to support cash application and accounts receivables reconciliation processes. To accurately match open receivables with payments and remittance information, data needs to be consolidated from multiple sources (including emails and web portals) and automatically isolated and deciphered (to facilitate a match regardless of the original format or structure). This process should be self-learning so that match rates improve over time. Over one-third of respondents have explored the use of ML/AI for cash forecasting and almost a quarter are looking to these technologies to improve payments and trading controls. For corporates with growing payment volumes or adopting instant payments (which are typically immediate and irrevocable) ML and AI are especially relevant. Generally, access to clean and standardized data is critical to implementing ML/AI technologies. Whereas currently we find fewer treasury organizations are actively considering the use of ML and AI to support investment or funding decisions, this use case may evolve over time as access to relevant underlying data elements improves.