Every so often, a once-in-a-generation technology comes around that divides opinion across the business and academic worlds. It’s hard to think of a recent technology that encapsulates this split opinion more than artificial intelligence (AI). To some, it represents everything from driverless cars for the masses to finding a cure for cancer. To others, such as Stephen Hawking, it could spell the end of the human race as we know it.
The debate is music to the ears of academics and philosophers. But for corporate treasurers immersed in the day-to-day pressures of reducing costs and boosting the top line, does AI really have anything to offer?
The truth is that true artificial intelligence is currently pretty far removed from corporate treasury. No finance officer today has the budget to overhaul their technology stack by splashing cash on AI. And treasurers would be wise to proceed with caution; it wouldn’t be a good idea to drink the AI Kool-Aid too early. Nevertheless, companies already have the opportunity to cherry-pick certain aspects of AI that can drive cost reductions and revenue growth.
Predictive Analytics in Treasury Forecasting
The adoption of predictive analytics is one way AI can assist corporate treasurers today. The term “predictive analytics” encompasses a variety of statistical techniques that assess current and historical facts to make predictions about future events. CFOs and treasurers can use these technologies to more accurately predict likely outcomes, from big geopolitical events such as Brexit to more micro events in a particular industry or locale.
Historically, using predictive analytics tools required advanced skills, as did understanding the results they delivered. However, modern predictive analytics tools are no longer restricted to IT specialists. They are becoming more intuitive, in order to encourage adoption by nonexpert users. As more organizations incorporate predictive analytics into operational and strategic decision-making processes, business users are becoming the primary consumers of the information. And business users want tools they can use on their own.
Predictive analytics encompasses statistical techniques including predictive modeling, machine learning, and data mining. In business, predictive models identify risks and opportunities by exploiting patterns found in historical and transactional data. These models capture relationships among many factors to allow assessment of risks, or potential risks, associated with a particular set of conditions.
They enable a CFO or treasurer to better predict likely outcomes from events such as FX moves, weather and natural disasters, or geopolitical shifts. The insights these models provide can help guide decision-making and preparation for the possible events, as the finance team might choose to use dynamic hedging to mitigate the event’s possible outcomes. Predictive models can now also calculate potential impacts from black swan events and other unlikely scenarios.
Machine learning is often confused with artificial intelligence. AI is the broader concept of machines being able to carry out tasks in a way that we would consider “smart.” Machine learning is a current application of AI based around the idea that we should really be able to just give machines access to data and let them learn for themselves, without explicitly programming the decision rules they will need to apply.
Cash forecasting involves formulating quantitative predictions for cash inflows and outflows in future periods. Machine learning enables an application to spot any payments that deviate from the norm in the payment flow. The payments can then be checked for fraud or mistakes before they are actually paid.
Technology vendors are also starting to put machine learning to work in predicting future cash flows. Efficient management of accounts receivable tends to be more challenging than managing accounts payable because receivables are affected by variables beyond the firm’s control. However, predictive algorithms can also benefit payables. Machine learning tools can use data from various systems to generate forecasts for incoming cash flow, sales, or even demand.
Such forecasting requires evaluating a large number of internal and external variables; weighting those variables; structuring an analysis of historical patterns; and incorporating real-time data from the business, procurement, and elsewhere. A treasury function could use regression calculations to determine the extent to which specific variable factors, such as interest rates, FX, or the price of a commodity influence movements in the price of an asset. In receivables management, the impact of early-payment discounts offered to customers might also lend itself to regression analysis.
Many companies still create their cash forecasts using spreadsheets and macros. These tools will likely be replaced by machine learning algorithms and other sophisticated AI tools. An effective, well-structured cash flow forecasting process improves confidence in forecasting data and better equips the treasury team to deploy dynamic hedging to protect against undesirable outcomes.
Time Series Models
Some sets of data points measured over time may have an internal structure—such as autocorrelation, trend, or seasonal variation—that any attempt to predict future values will need to account for. As a result, standard regression techniques cannot be applied to time series data.
Time series models have been developed to decompose the trend, seasonal, and cyclical component of the series. Modeling the dynamic path of a variable can improve forecasts, since the predictable component of the series can be projected into the future.
The Role for Robotics
Then there is robotics—which is, essentially, letting machines do the work of humans. We have yet to see robots sitting behind the treasurer’s desk making strategic decisions, but automated processes are available for many treasury tasks, from cash forecasting to reconciliation, and from cash sweeps to currency risk management. Treasurers with access to rules-based automation and exception-handling technologies now have the ability to remove a whole series of manual and time-consuming tasks from their day to day.
Treasury tasks likely to benefit from robotics include document processing, invoicing, approving transactions and identifying least-cost payment methods, and providing regulators with the required data to automate compliance. For example, the term “smart payment” refers to the automation of processing for payments and remittances that are currently paper-based. Smart payments improve the speed and efficiency of payments, with parameters on what gets paid and what requires higher-level authority review before remittance.
Replacing human labor in high-frequency tasks can lead to a significant cost reduction for some activities. Moreover, because robotized tasks can move to times when the company is consuming fewer IT resources (e.g., overnight or weekend), they can contribute to the optimization of finance processes. At the very least, automation will help treasury staff refocus to become better decision-makers and proactive risk managers by reducing manual tasks.
AI in Cash Forecasting
An effective, well-structured cash flow forecasting process improves confidence in forecasts and helps to deliver more accurate results. Forecasting accurately can be complicated. Firms must synthesize and correctly balance both internal and external variables, and they must take into account historical patterns, real-time data from their business operations, and uncertainty about how well business units themselves can access their own current status. Collecting reliable, sufficient, and relevant data is paramount in building an efficient forecasting process. Historically, collecting the data and then pulling it together on a single platform for analysis have presented significant challenges.
Multifaceted calculations involve more data than any human can accurately model, even with the aid of tools such as Excel. For firms with international operations and supply chains, AI presents an appealing solution for financial planning and treasury teams. Given the rapid growth in corporate data volumes, AI procedures are not just useful for providing new insights, but necessary to synthesize all of the relevant data so that firms can compete effectively.
AI-centric solutions can incorporate systems that parse myriad information sources and calculate external political and economic risks, systems that reveal insights from a firm’s own data, and systems that construct models of financial information and inputs. The information provided by an AI solution can then be studied, and the impacts of each system’s suggestions for the business can be deduced.
From everyday forecasting to the more glamorous end of robotics and predictive analytics, adopting specific aspects of AI can certainly give a big leg up to treasurers trying to navigate these cost-conscious times. Those that take a thoughtful, focused approach—instead of allowing their budgets to swept up by all the hype surrounding AI—will be the ones that benefit the most in long term.
Mark O’Toole is the vice president for commodities and treasury solutions at Openlink. O’Toole currently leads global sales, strategy, and product marketing for Openlink's Commodities & Treasury business and has worked in a number of pivotal roles within the company. He has been with the firm since 2006 and is based in New York.