AI Makes Cash Forecasting Less Art, More Science

By January 16, 2020January 24th, 2020No Comments

HighRadius says AI continually improves cash-flow forecasts.

Artificial intelligence (AI) may “take the art” out of cash forecasting while perpetually improving forecasts, but there will still be plenty of work for treasury staff.

That was the message imparted by Tracey Ferguson Knight, director of solution engineering for treasury at HighRadius, which hosted a recent NeuGroup meeting. She noted that cash forecasting today can be hobbled by inefficient technology—think Excel spreadsheets—and difficulty gathering and leveraging the correct data sets, an art often mastered by one person whose departure can put the company at risk.

Accounts receivable (AR) tends to be the source of a large amount of uncertainty and volatility in a company’s cash forecast, Ms. Knight said, and companies typically take the average between invoice dates and when payments arrive to generate forecasts across AR. A better approach, she said, would be to calculate an average for each customer and improve accuracy by including pertinent variables, such as a customer taking longer to make larger payments.

HighRadius has identified more than 60 such variables at the invoice and customer level, although typically only a portion will be correlated to when a customer pays. Nevertheless, introducing those variables and doing so for thousands or even hundreds of thousands of customers is a gargantuan task, especially given the antiquated technology companies often use.    

Enter AI. Once the variables and appropriate algorithm are chosen, the AI engine crunches that data and provides ongoing forecasts, tapping data sources approved by the corporate, such as its enterprise resource planning (ERP) system and banks. Manually gathering and consolidating data and manipulating spreadsheets is no longer necessary.

The algo learns. The algorithm may look at customers’ two-year history in monthly, quarterly and annual increments and, for example, recognizing increased sales over the last two quarters, perhaps from a new product launch, and incorporate the uptick into new cash forecasts. “The algo is constantly updating itself and becoming more accurate,” Ms. Knight said. HighRadius aims for a 30% improvement in cash forecasting accuracy off the bat, with continual improvements as AR history grows.

Humans still vital. Ms. Knight said it’s a misconception that AI will replace humans, since people must figure out the best variables and algorithms to use and adjust them to account for impactful events and changes to data sources. “It’s people utilizing machines to get the most value, but people are still very much involved,” she said.   

Dependability. HighRadius’ system calculates the variance between forecast and actual cash flows over different time frames. Drilling down by category, subsidiary or region can reveal where the variances lie and clearly identify the greatest areas for improvement.

The treasurer version. HighRadius has taken its years of experience forecasting AR and applied its methodology to the other cash forecasting components, including AP, payroll and expenses. Ms. Knight said that enables today’s “digital treasurer” to make informed decisions without having to gather, calculate and present data.

Please join treasury and A/R leaders at Radiance 2020 in Dallas Feb. 5-7. Get 30% off when you register using code rad2020neugroup. And inquire about your eligibility to attend the NeuGroup treasurers’ meeting before the event to share AI forecasting experience and get cash feedback from peers.

Antony Michels

Author Antony Michels

More posts by Antony Michels