Machine Learning & Treasury: A Bright Future Emerges
Over the span of a generation, treasury organizations in many large companies have migrated from a reliance on spreadsheets toward the use of treasury management systems.
Over the span of a generation, treasury organizations in many large companies have migrated from a reliance on spreadsheets toward the use of treasury management systems. These
So extraordinary is the digital transformation of treasury that many believe treasury teams of the future will primarily be engaged in value-added activities in areas like optimizing cash flows and currency management, instead of routine transactional tasks. Datadriven insights that staff acquire from analytics tools will also be increasingly valuable to other parts of the enterprise.
Expectations are that within five years, most treasury management systems will include embedded RPA—which automates repetitive processing tasks using rules-based logic. In fact, some vendors have already started to integrate these technologies. Robots automatically access, assess, and process huge data volumes involving cash positions, interest rates, payables, receivables, and foreign exchange rates, performing these tasks faster and more accurately than humans could.
Algorithms in RPA tools can also instantly highlight any variances from expectations, illuminating a path toward reconciling issues. When RPA is further augmented with data analytics, treasury can develop more transparent and precise forecasts of long-term liquidity.
Within 10 years, treasury platforms will “look and act like no treasury management system today,” says Bob Stark, vice president of strategy at Kyriba, a provider of a cloud-based treasury management system.
Intensifying Pressures
The future may feel a bit scary for treasury professionals, as their roles will have to change. But the consensus among many treasury experts is that the function is headed for its heyday. Just in time, too, given the recent regulatory upheaval increasing the compliance obligations of finance.
Several new rules—among them Basel III, Dodd-Frank, IRS Section 385, and changing Public Company Accounting Oversight Board (PCAOB) audit standards— exemplify the increasing and complex pressures on many treasury organizations. In some cases, these regulations compel corporate treasury groups to take on additional responsibilities, some of which were previously assumed by banks.
“We’re at the threshold of momentous changes ahead,” Stark says. “Chief among them are the responsibilities of treasury organizations and the roles played by treasury staff. The current paradigm of treasury reconciling bank accounts will shift more and more toward strategic and analytical decision-making. Treasury must again rethink where it can best provide service.”
He is referring to the paradigm shift of the 1990s, when treasury workstations first came on the scene. At the time, these systems were purchased almost solely by very large organizations. Gradually, treasury organizations at midsize and smaller companies began shifting from spreadsheetbased cash management, reporting, and forecasting processes to using the automated platforms.
“Treasury management systems resulted in the need for treasury to relinquish certain processing responsibilities, once more efficient tools were available,” Stark says. “New roles materialized within treasury, increasing the strategic importance of the function. We’re now at a similar juncture.”
Other treasury experts agree that the function is now at a crossroads—one posing momentous changes in the services provided by, and capabilities of, treasury staff.
“On the one hand, technologies like RPA and machine learning are re-imagining the capabilities and purpose of treasury, but on the other hand, they also necessitate the retooling of the competencies of the people within treasury,” says Michael Kolman, co-COO, head of business development at ION Treasury, which comprises a portfolio of seven branded treasury and risk management solutions including Reval, IT2, and Treasura. “Skill sets have to change, and will change.”
Technological Upheaval
Core business functions in enterprises across every industry are confronting the opportunities and threats posed by digital transformation. Once information is digitized, a computer can capture, access, search, exchange, and integrate different data elements. Corporate functions that have digitized a significant portion of their data will then use this information to optimize business and operating workflows. The final step in a company’s digital transformation is to reorchestrate the organization’s digital footprint, internally and with third parties, to create a collaborative transactional digital ecosystem.
In this journey, RPA is a critical technology because it can liberate staff from performing routine and repetitive high-volume, lowcomplexity treasury tasks. When robots execute these tasks instead, the result is more cost-effective processing.
RPA also enhances accuracy, since robots do the same things in the same way every time. For compliance purposes, this leap forward in precision reduces the risk of the dire consequences that would result from an accounting mistake. RPA tools perform processing tasks at breakneck speeds on a continuous basis, in a working timetable that human beings could not match. The robots also are better at organizing data to create reports. These various enhancements in productivity, accuracy, and compliance add up to significant treasury cost savings. Further, because automation reduces the chance of human error, RPA tools improve confidence in the veracity of cash forecasts.
Both customers and vendors of treasury technology platforms benefit. “We’re using RPA to automate tasks that we had previously offshored to save costs, and we are saving more in the process,” says ION Treasury’s Kolman. “It makes no sense anymore for human beings to reconcile a bank statement or execute a payment over and over again, when robots can be designed to do the same things more rapidly and easily.”
Another plus is that RPA presents the opportunity for system users to effortlessly upgrade the product. “Every software system needs to be periodically replaced with a new version of the same product, on a three- to five-year basis in many cases,” Kolman says. “Robotics gives us the ability to reduce the typical number of steps involved in upgrading the system, making the upgrading process much faster.”
He explained that the robots simply perform the repetitive steps previously performed by humans. If there were previously 20 steps involved in upgrading the systems, there may be 5 steps now because the robot has already performed the other 15.
In addition to these benefits, RPA facilitates easier communications with other bestof-breed systems within treasury, such as currency management tools. Data can be captured and transferred from one system to another with ease, permitting the integration of different data elements via application programming interfaces (APIs).
Different functions within a business, like financial planning and analysis (FP&A) and supply chain management, could certainly benefit from betterinformed cash flow and cash forecasts to support their decisionmaking. Small wonder that the buzz around RPA within treasury organizations is amplifying.
“In just the last three weeks, I’ve had multiple conversations with clients and potential clients about RPA, wanting to know how it will affect their financial business processes, and—by extension—their roles,” says Andrew Gage, vice president of strategic market development at FiREapps, a provider of FX exposure management software. “It is our belief that robotics, machine learning, and predictive data analytics are altering the competitive landscape for treasury organizations. Data is king, and whoever has control of it to produce enterprise value at the end of the day is the winner.”
Next-Stage Machine Learning
In a way, RPA is the foundation for the implementation of other technologies that may culminate in increasingly “smart” treasury systems. Chief among these tools is machine learning—software that enables computer systems to actually learn from data they encounter. An example of machine learning is the spam filter in your computer, which has been taught to ferret out and flag unwanted email.
“Machine learning takes your existing technology stack in treasury to the point where people become involved even later in the process than they do presently with RPA,” says Kyriba’s Stark.
For example, take the routine treasury task of capturing the company’s cash position. Downloading details from each bank individually is timeconsuming, and many treasury teams are already moving to automate that process. What machine learning adds to the equation is the potential to quickly root out the causes of any variances between forecast and actuals.
“Typically, an experienced cash manager is able to dig through the details to unearth the specific reason or reasons” for the variance, Stark says. Machine learning can convert that human research and decision-making into an algorithm that does the chasing instead, culminating in a level of automation where answers are nearly immediate.
Thus, machine learning may accelerate cash forecasting processes and help guide more accurate and more informed decisions. “When you have a better grasp of excess liquidity, you may decide to pay down the credit line,” Stark says. “That’s just one decision among many.”
ION Treasury’s Kolman also extols the enhancements in cash forecasting, which he says are “top of mind” for a lot of the firm’s treasury clients. “Certainly, cash flow is an organization’s lifeblood, and cash forecasting is a vital core functionality for treasury operations,” he explains. “It’s also one of their biggest challenges. Treasury must accurately analyze cash flow, deliver accurate cash positions, and make accurate cash forecasts, while discerning variances and the reasons for them. Technology provides this heightened level of accuracy and assurance.”
Yet another benefit of machine learning is the technology’s ability to quickly and efficiently identify possible instances of fraud. For a treasury organization to detect fraudulent activities today, a human must screen every payment. However, a rulesbased engine could be designed to look for specific patterns that raise red flags, such as a customer opening a new bank account, billing for certain services that are not appropriate for a particular account, or workflows that are atypical in some way.
Machine learning further helps a company optimize cash management by identifying how much working capital the organization needs. “You’re better able to line up different choices regarding excess cash, such as investing it in money market funds; giving it back to the credit line; increasing longerterm debt; or doing something more outrageous, like working with procurement to figure out discounts with suppliers,” Stark says. “People make the ultimate decision, but the analytics line up the recommendations.”
Treasury Burdens
Several surveys indicate that treasury teams can expect their cash forecasting and cash management processes to come under increasing scrutiny, making the benefits of machine learning even more important. For example, in a 2016 survey by SAP, more than three-quarters (76 percent) of finance and treasury respondents projected that cash management would become more challenging for treasury staff within the next five years, with one-third (34 percent) citing a need to improve the accuracy of cash forecasting and another third (33 percent) stating that accuracy, quality, and consistency of cash flow data are paramount considerations.
A 2018 survey by the Economist Intelligence Unit (EIU) of 300 senior corporate treasury executives across multiple industry sectors echoes the same themes. According to the EIU survey, more than 55 percent of respondents are experiencing “knock-on” effects on treasury from “sector disruption.” Respondents perceive technology as an effective tool for limiting this impact; 56 percent cited data analytics as beneficial for the treasury function.
ION Treasury’s Kolman concurs. “The combination of RPA, machine learning, and bigdata analytics offers the ability to produce more accurate forecasts automatically,” he says. “We’ve invested in machine learning to support multiple parts of our portfolio and are beginning to see benefits.”
It Just Gets Better
Down the line, further enhancements in these disruptive technologies are likely to combine in unique ways to dramatically alter the treasury function—and the treasury profession. “Technology will do much of the heavy lifting in treasury in the future,” Kolman projects.
This is a development that few would have imagined a decade ago, he adds. “After the financial crisis in 2007–08 when companies struggled to raise capital quickly through traditional means, there was a heightened focus put on cash forecasting and companies established 13-week, 26-week, and sometimes even longer direct cash forecasts which required significant effort and sometimes limited accuracy,” Kolman says. “Forecasts were built using spreadsheets which was the quickest way to implement.” “But we were still building out the forecasts using Excel, which remained the quickest way at the time.”
Fast-forward to today’s digital transformation: “Just a decade later, we now have the opportunity to produce extremely high-quality, accurate, and near–real-time forecasts. It’s a pretty astonishing development,” Kolman says.
Still, whether corporate treasuries will invest in more robust systems that embed these disruptive technologies remains to be seen. Treasury management systems have been around for almost 30 years, and their market penetration is still nowhere near 100 percent. The experts we spoke with are, nonetheless, sanguine about prospects for future market growth.
“The future bodes well for more rapid implementation,” Kolman says. “Treasury organizations within companies that are historically risk-averse when it comes to adopting technology are now participating in treasury management system pilot programs involving machine learning and other automated solutions. No longer is treasury ignoring their potential. And that makes for very exciting times.”
These “exciting times” are not happening overnight. Stark notes that Kyriba, for example, is taking an “evolutionary—and not a ‘Big Bang’—approach” to building robotics and other analytical tools into its cloud-based treasury management system. And Kolman says ION Treasury has plans to invest incrementally in machine learning, given the technology’s potential to achieve faster response times, lower total cost of ownership for clients, and ability to automate testing. “Our mission is to enable automation through innovation, but we pursue this mission thoughtfully,” he says.
Tomorrow’s Treasury Professionals
Measured approaches aside, corporate treasury will, in the future, be a very different organization. Expectations are that treasury will become more of a strategic resource—not just to colleagues in finance, but throughout the enterprise. (See the sidebar, Treasury at the Center)
Treasury at the Center
As the treasury organization begins to reap the benefits of sophisticated technology tools like robotic process automation (RPA), data analytics, and machine learning, it is poised to become a centralized hub of financial insights for the rest of the business. Corporate treasury functions that are able to move into this role may improve cash forecasting and optimize liquidity management across the enterprise.
In the very near future, treasury teams can expect to hear from staff in other finance groups, such as financial planning and analysis (FP&A) and accounting, who are looking to learn from treasury’s new, analytics-driven insights on cash flow and forecasts. Treasury teams will likely start to receive similar requests from groups outside finance, as well—from business unit leaders to compliance and audit, among others.
“Treasury already is receiving emails from their business colleagues in other parts of the enterprise regarding cash flows and working capital,” says Michael Kolman, coCOO, head of business development at ION Treasury. “With treasury’s role in the future increasingly transforming into a cash optimization analyst, reliant on RPA and machine learning to make data-driven decisions, internal customers will reach out even more with their questions and concerns.”
As this occurs, Kolman says, treasury staff have the opportunity to rethink their role to better serve these stakeholders. It’s an interesting concept, and some treasury groups are already reconsidering how they can support the broader organization in areas like currency management. As new systems focus increasingly sophisticated business intelligence capabilities on foreign exchange (FX) data, the resulting insights are relevant to—and being requested by—a growing audience within the enterprise, says Andrew Gage, vice president of strategic market development at FiREapps.
“When I started a decade ago in the burgeoning FX technology space, the system of choice for understanding currency risks was still a 1,000-page spreadsheet,” Gage notes. “Each one of the cells in that massive document contained a calculation that was prone to errors. Now we’re at a point where modern treasuries are able to manage currency in a highly automated fashion at amazing speed, sourcing data out of the ERP [enterprise resource planning] system to make rapid, informed, and fact-based decisions. And we’re just beginning to tap what treasury will be able to do in the near future.”
By connecting and integrating modern FX management systems with other bestof-breed finance systems and applications, Gage says, treasury will be positioned to routinely pass on information about currency exposures to other parts of the enterprise where this information can improve business decision-making.
“As data is mined out of the ERP and other source systems, it can be pulled into useful insights for FP&A down the hall—assuming this information is presented in an intuitive way,” says Corey Edens, FiREapps’ chief solutions officer. “FP&A can more quickly calculate the transactional impact of currency risks for reporting purposes in the 10-K and 10-Q.”
Other areas of the business also may find this information useful in calculating the currency impact to their revenue and cost of sales. “A case in point is the use of currency data in the supply chain to make sourcing decisions,” Edens says.
A wide range of stakeholders are “hungering” for these data-driven analyses, Gage says. He’s not alone in this perspective. “By digitalizing the responsibilities of treasury managers, the universe of applications elsewhere in the organization dramatically expands,” says Bob Stark, vice president of strategy at Kyriba. “Treasury is in a position to provide rapid information on cash to anyone in the business who needs these insights for decision-making purposes. Leveraging RPA and machine learning, treasury can quickly isolate data to provide useful analyses responding to what these stakeholders need to know. The potential is huge.”
Gage agrees, noting the exponential impact of this advice. “As different parts of the enterprise take advantage of treasury’s analyses, they’ll begin to ask for more things beyond just their exposure in currency risk at the corporate level,” he says. “They’ll want to know the currency risk by line of business, by geography, by customer, and so on. They’ll want other data mined for information on what a particular currency exposure is doing to their monthover-month income statement. Once they grab these insights, they’ll want more and more. And what’s great about these tools is that they will comply with these demands, doing more of the digging as needed.”
Treasury now stands at an inflection point, where staff can leverage systems that are cloud-based, have more processing horsepower, and include evermore-advanced analytics. Gone will be the days of treasury staff spending hours each month reconciling bank statements. In the future, they will be better aligned with the rest of the enterprise, providing stakeholders throughout the company with valuable insights—to the betterment of the bottom line.
Tomorrow’s Treasury Professionals (cont’d)
“To expect a treasury analyst in the future to do nothing more than reconcile accounts is reductive,” Stark says. “Future treasury roles will require traditional expertise, in addition to a range of newer analytical skills and decisionmaking abilities.”
Soon, treasury professionals will be expected to know how to effectively leverage RPA, data analytics technologies, and machine learning tools. This will likely mean companies need to provide training for their treasury staff if they want to get the most out of these technologies.
“We’re getting to a point where the volume of data coming into the treasury organization, from different internal and external sources, is growing at a blistering pace, with millions upon millions of rows of highly complex data,” says Corey Edens, FiREapps’ chief solutions officer. “Capturing, harvesting, and integrating this disparate data to tell an analytical story within treasury—and across the business—requires an intuitive use of highly sophisticated technology. People have to be adept in using these tools, which will require hiring individuals with broader skill sets, in addition to retraining current staff.”
Kolman agrees that the technologies are becoming increasingly sophisticated, pointing to ION’s ongoing development of its Treasury Anywhere tool, which is being designed for users’ interactions on a mobile device. “Right now, we’re experimenting with biometrics and voice-activated software to use the tool,” he says. “You can ask ‘What’s my cash position in Europe?’ and the device will answer back. This stuff is no longer science fiction; it’s real, it works, and it’s coming.”
Science Wins the Day
The shift in status of the corporate treasury function from that of a technical resource to more of a strategic business partner is consistent with the predictions of the 2016 SAP survey on the future of cash management. Nearly two-thirds (62 percent) of respondents said that treasury must expect to take on higher-value activities in the future, providing deeper and faster insights on cash positions and forecasts to senior management and other parts of the enterprise. This shift is now occurring, the interviewees confirmed.
Meanwhile, treasury technology platforms will do more and more of the menial, manual work that people previously had to perform. Within the not-too-distant future, treasury organizations will shrink, as both automation and analytics technologies absorb much of the current treasury workload. Some people will be displaced. Others will need to add value through their skilled use of technology to analyze and interpret the results of the data analyses.
The rest of the enterprise will call upon these highly trained analysts to provide advice on their own needs for strategic cash management. In the SAP survey, 82 percent of respondents predicted that non-treasury staff would reach out to treasury in the future to better understand their cash positions. “One can easily imagine treasury having a seat at the strategy table to assist complicated concerns and deliberations over more complex risks and regulations,” says FiREapps’ Edens.
Ten years from now, spreadsheets will likely have gone the way of the abacus; young treasury professionals will read about Excel in their college history-of-finance classes. “Treasury departments should create their future vision today where repetitive tasks are automated and liquidity forecast accuracy improves, because with new technology the vision will be achievable,” Kolman says.
Stark concurs: “Now is the time for treasury to re-imagine where they can provide the most value tomorrow.”
Also from the November 2018 Special Report:
SPONSORED STATEMENT – FiREapps: Combining BI Tools & Enterprise Currency Analytics
SPONSORED STATEMENT – ION: Seven crucial tips for selecting the best Treasury Management System for your business
SPONSORED STATEMENT – Kyriba: How Technology Creates Intelligence in Treasury