It's difficult to overstate the volume of buzz around artificial intelligence (AI). Thesetechnologies are finding their way into everything from thesmallest consumer devices to the largest supercomputers—essentiallyinto every corner of enterprise software and the farthest reachesof the cloud. And so journalists, analysts, and corporatedecision-makers are all spending a lot of time writing and readingabout AI, innovating around the concept of artificial intelligence,and integrating it into products.

The hype isn't necessarily overblown. These technologies andtheir various components—like deep learning and machine learningalgorithms and robotic process automation (RPA)—hold promiseto accelerate processes, reduce errors, improve efficiency, andlower costs, particularly for areas of the business that are highlyautomated and data-rich. That would include the complexorder-to-cash cycle, a vital part of any company's operations. It'swhere businesses interact with their customers, orders are made andfulfilled, bills are sent, disputes are resolved, and payments arereceived and processed.

A company's reputation may rise or fall based on how it managesits order-to-cash cycle. Over recent decades, many companies haveworked hard to improve operations in this area. Some have leveragedcustomer order management software to make them moreefficient. AI and machine learning further extend the benefits ofautomation. The key to maximizing the benefits is to leverage theright AI solutions, in the right areas of the order-to-cashprocess.

Benefits of Order Processing Automation

Many businesses have already put a lot of effort into automatingtheir order-to-cash activities. This makes sense, as the cycle iscomplex. Orders may come in from multiple sources—includingdisparate point-of-sale systems, fax, email, the company website,and/or electronic data interchange (EDI) connections to businesspartners. When processes such as creating invoices, ensuringpayments come in on time, and resolving disputes involve extensivemanual work, problems can range from slow order fulfillment orincorrect shipments to data-entry errors and poor customer service.Staff who should be working closely with customers and partnersinstead spend their time doing the grunt work of receiving orders,shipping products, and processing payments.

Order processing software currently on the market can address alot of these challenges. By automating many of the manual steps inthe order-to-cash cycle, these solutions can save companies moneythrough streamlined order processing, improved employeeproductivity, and reduced equipment and personnel spending. Theymay also enable customer service representatives to focus more timeon ensuring customers' needs are met. In addition, orders may beprocessed and fulfilled more quickly and with fewer errors.

Order-to-cash automation has been a boon for many companies, butthat doesn't mean there isn't room for additional improvement.Businesses that automate certain processes, such as orderdocumentation, collections, or dispute resolution, sometimes leaveother activities in human hands. For example, staff may continue tomanually compare invoice numbers against open orders or matchremittances to payments to ensure that what comes out of theautomated workflow has been done correctly.

AI and Machine Learning WithinOrder-to-Cash

Artificial intelligence and machine learning can improve theorder-to-cash cycle in myriad ways. In business, data is the coinof the realm, and these technologies are designed to leverage datato improve business operations and decision-making. They do that byanalyzing vast volumes of information, looking for patterns thathumans could not be expected to detect. When the AI lens is focusedon order-to-cash data, businesses can leverage detected patterns tostreamline processes and save time.

Here are a few examples of areas in which AI technologies mightbe able to further extend the benefits of a more traditionalorder-to-cash automation solution:

Order processing. Many companies stillsend in their orders via email. Traditionally, vendors have paidsomebody to sort through the emails and pull out details such asorder numbers and customer identifiers, then route the emails tothe appropriate destination. Much of that work can be doneautomatically through an AI engine.

The engine can automatically identify the data in the email;recognize which items the customer is ordering; determine whethermultiple orders in a single email need to be processed separately;ascertain whether there are duplicates; and, if so, wait for ahuman to pick the right order and process it. The AI engine canalso resolve issues in how data is presented. Are dates from thecustomer's enterprise resource planning (ERP) system written withdots rather than dashes? Are zeros kept in? The software canautomatically format data discrepancies and move orders forwardthrough the process. Thanks to machine learning, the AI engine willalways remember changes that are made and can apply them to futurecustomers as well.

Dispute resolution is an importantcomponent of any company's order-to-cash process. Resolving paymentdisputes quickly and efficiently is critical in ensuring thataffected customers remain satisfied with the organization. However,when the process depends on one-off, manual consideration of eachcustomer dispute, it is time-consuming for staff, which can place adrag on resolution for customers and add to costs for thebusiness.

Most disputes don't require any correction. If customers arecalling a customer service representative (CSR), they may have beenbilled for something they didn't order, or they may want to reportthat the order is incorrect. The CSR may have some informationabout the sale—the product, amount, cost, and other data—andrecords of prior complaints, but he or she doesn't necessarily havethe authority to make a final decision. There's typically anapproval process that kicks in after the complaint is recorded.Resolution may require further discussions with the customer,gathering of more information, and documentation of steps the CSRis taking. The process may take three days to two weeks, duringwhich time the customer may be placing additional orders.

AI technologies could greatly accelerate dispute-resolutionprocesses. If software specifically designed to recognize patternsin data were unleashed, it might be able to rapidly identify whichcustomer concerns are most likely to be valid. A business that canbetter prioritize disputes, based on which are most likely to needhuman review, will speed up the time to resolution for legitimatecustomer complaints.

Invoicing.  In accountsreceivable (A/R), machine learning ensures that invoices reach therecipients more quickly and error-free. Invoicing traditionally hasbeen a manual and painfully slow process, with a paper invoicemoving from hand to hand internally before being sent out to thecustomer, who then turns it around and sends it back in withpayment. At the least, automating the process of deliveringinvoices means customers get them more quickly and can turn aroundpayment faster.

But AI also streamlines compliance with disparate regulations.Laws around invoicing and payments vary from one region to thenext—special stamps are needed in some, digital signatures orparticular layouts for others—and in many jurisdictions, theychange frequently. Manually keeping up is almost impossible.Automated cash collections also are quicker and easier, and AIengines will remember who at a particular customer company is bestto contact, or if they prefer to be contacted via phone oremail.

AI also can play a crucial rolein detecting fraud or deceit in invoicing, because detectingpatterns and anomalies is a key capability of machine learning.Typically, when an invoice is sent, it's printed and then put intothe mail. No one is looking at the data on the invoice. Byincorporating AI into the process, the engine is analyzing all thedata as it's going through and is detecting invoices on which orderfrequency or amount, for example, fails to fit the establishedpattern for a certain customer. When it detects a potential issue,the invoice can be flagged for staff to examine. For thosereceiving invoices, AI can be used to ensure that nothing is out ofthe ordinary—that they are being billed the same amount for thesame services they usually use. The AI engine also can verifycustomer and invoice data and banking details before allowing apayment to be made.

These are certainly not the only possibilities. AI and machinelearning technologies are designed to take in information and learnfrom it. Thus, they can enable individual processes to learn fromone another and automatically improve their performance over time.Essentially, any process that relies on human investigation of datais ripe for AI-driven efficiency improvements. The potential for AIimprovements is limited only by the company's development resourcesand analysts' imagination.

Considerations in Planning

AI technologies have been around for decades, but they'veexploded in recent years as new techniques like deep learning andneural networks have been improved and applied. Machine learningessentially uses algorithms to analyze and learn from data and thenmake decisions based on what it's learned. Deep learning structuresdata in layers, helping to create neural networks, which aresystems designed to reflect neuron patterns of the human brain.

In many ways, deep learning is about using images and data toidentify information—maybe identifying and remembering acustomer—and then using that information in the order-to-cashprocess. This may include on-boarding a new customer and accuratelyextracting the data to automate and improve the process. Deeplearning can also help avoid downstream problems by ensuring thatthe data used in invoicing is correct so that there are no problemswith billing. The advantage of deep learning is that it can usealgorithms and vast amounts of data to teach itself. It doesn'tlearn simply through one or two orders, but by seeing millions oforders, images, and bits of data.

The potential benefits are significant, but businesses shouldn'tbe in a rush to adopt the first AI-based technology they can gettheir hands on. Treasury and finance managers should carefullyconsider their options. Four key considerations should help guidethe selection process.

1. Should you look to thecloud?  Cloud providers like Amazon WebServices (AWS), Microsoft Azure, and Google Cloud Platform are onthe cutting edge of AI and machine learning development. Thesevendors offer a broad range of services that companies can leverageto run many of their order-to-cash processes—think order receiving,invoice creation and sending, and payment collection—on a publiccloud infrastructure. Customers using such solutions can host theirdatabases in the cloud or link these processes back to their owndata centers, and they can leverage the cloud provider's deep AIcapabilities to automate the order-to-cash cycle.

Using a cloud-based AI provider enables a company to leveragethe convenience, scalability, and subscription-price models thatthe cloud is known for, and the burden of expertise in areas suchas infrastructure, AI, high availability, and data protection andretention lies with the provider. However, some companies areuncomfortable letting such mission-critical processes and data,including customers' personal information, run outside of theiron-premises environments.

A growing number of vendors of order-management software areincorporating more and more AI into their products. Companiesconsidering how to leverage AI to improve their order-to-cashprocess should weigh their priorities and then answer the questionsthat will arise: What functionality is most critical? Which cloudproviders—or on-premises software vendors—do the best job ofproviding those capabilities? To what degree should we be worriedabout security, and does each vendor satisfy our concerns? Whatservices and support come with each option?

2. How can you ensure your approach is appropriatelyholistic?  AI enables companies to take aholistic view of their business processes. The technology can drivedata analytics, to show where the system is performing well andwhere it needs improvement, and can then automatically make changesbased on the data to drive those improvements. AI solutions canaddress the entire order-to-cash cycle, from the moment an order issent through to when it is fulfilled, invoices are sent, billing iscompleted, and payment is taken. At each point, the process isautomated, which accelerates activities and reduces errors.

In addition, the automation available through AI bringsconsistency to the data and procedures in each step of theorder-to-cash process. The data that is gathered from the initialorder is the same as when the order is fulfilled, the invoice issent, and payment is made. Up and down the process, each step isoperating from the same data, rendered in the same way and taken orsent out through consistent procedures. Moreover, each invoice issent to a particular customer in the same way.

The data connected to each customer is also collected andremembered by the AI engine. For example, if a customer thatnormally buys 1,000 widgets puts an order in for 100,000 widgets,the system automatically detects and flags that order. At the otherend, if an invoice has anomalies, the system will detect a changein patterns and alert the company.

When AI and machine learning algorithms span multiple systemsand departments, the order-to-cash cycle improves further, as thesystem learns and algorithms enable it to complete processes fasterand more accurately. This capability to learn creates anenvironment where, once the system is in place, it's pretty much ahands-off operation.

As your organization works on developing an AI-driven solutionfor the order-to-cash cycle, think about taking a holisticapproach; look for solutions that could tie together all processesfrom customer orders to final payments. AI-based solutions learnbetter and faster when they take in more data. That data mayinclude information that lets the solution better know thecustomer—not only identifying numbers but buying histories, moneyspent, and how they like their invoices sent. With this type ofinformation, the solution can pick out patterns that might indicatewhen something changes with the customer, if there's payment fraud,or whether an order's correct.

AI systems are as good as the data that goes into them, and themore data points they get, the more complete the view is of thewhole operation. Also, being able to see information from multipleother systems will enable the solution to determine how processescan best work together. This may be something as simple as ordernumbers. Suppose that when an order comes in, an order number isgenerated. Through AI-based automation, that number isautomatically put onto the shipping paperwork, the invoice, andeventually the payment receipt. For companies that are processingthousands of orders every day, automation of even a small step likethis can lead to significant efficiencies. It's a key benefit of AIand machine learning.

3. Don't rush the decision, or theimplementation.  Businesses may not want torush into adopting AI in one fell swoop. These solutions present alot of options and require decision-makers to make many discretechoices. They should take their time to assess their situation andthen evaluate which options best suit their needs.

Once a company has selected an AI-enabled order-to-cashsolution, it can take an incremental approach to deployment.Individually rolling out different parts of the overall systemcould save time and headaches by minimizing the impact of misstepsand mistaken decisions taken early on. It also allows the companyto develop expertise in AI and machine learning as the solution'sscope and sophistication grow.

For many organizations, the first step is to understand theircurrent processes, to get a better idea of how they can beimproved. Companies then need to set goals. Why are you automating?Why do you want to bring AI, machine learning, and deep learninginto your processes? What problems are you trying to solve?Companies need to understand what the data will show them and howthey can get that data. After that, it's a matter of implementingthe new process and bringing in the solutions that will enable themto leverage the AI capabilities.

Key to all this is getting end users involved at the beginningof the process. Some staff may worry that AI technology will resultin the elimination of their jobs. For most people, implementingthese solutions usually just means redirecting their efforts toother tasks. However, end users need to feel comfortable with theAI technologies and buy into the program, to ensure it runssmoothly from the start.

4. How will you measure the system'ssuccess?  Before you begin deployment,ensure that you will be able to track what data the solution isanalyzing, what types of trends it's looking for, and—just asimportant—what it is achieving. Make sure you have clearlyexpressed expectations of how the performance of order-to-cashprocesses should improve as a result of the AI solution. Keyperformance indicators (KPIs) are crucial to any businessoperation, and AI technologies can help customers gain greaterinsights into system performance.

With AI, the world of KPIs changes in tune with what becomesimportant. In the manual world, if a piece of data is extracted atone point and changed downstream because it's incorrect,performance metrics may not reflect that situation. With AI, notonly are those downstream issues eliminated, but the company's KPIshighlight steps that may improve the business or ways to movepeople around to better serve customers.

The metrics shift from measuring the number of orders coming inor the number of lines in the order to describing howorders are coming in, how many touchpoints there are, and who'stouching the order at each point in the process. From there, giventhe amount of data in the AI solution, companies can see if thereare problems with customers, a single employee, or the process.Correcting something downstream is no longer a problem.

Then the metrics can change again, to measure how much fasterthe company can get the product out the door. In other words: Isthe AI system saving or making us money? Is it really enabling usto better serve customers? The answers give businesses a betterview of the entire operation. They can educate their up-frontpeople, their CSRs, and show them how the company's activities atone point in the order-to-cash process affect other points.

AI Living up to the Hype

Businesses are catching on to what AI can do for them. Accordingto Statistica, 84 percent of enterprises say investing in AI will lead to competitiveadvantages, and 63 percent say AI will be needed in the future toreduce costs. Gartner analysts are forecasting that by 2020, AItechnologies will be in almost every newsoftware product and service.

Companies wanting to get a jump on the competition by reducingorder processing time, costs, and errors, while improvingefficiency and customer service, should give some thought to theways in which AI technologies could support these goals. In today'sincreasingly fast-moving and constantly changing business climate,automating order-to-cash processes usually translates to happiercustomers who are less likely to look elsewhere for suppliers.Learning to better leverage data to improve every step in theorder-to-cash cycle enables a business to reap the much-discussedbenefits of AI.

Eric Bussy isworldwide corporate marketing and product management director atEsker, a producer of AI-driven software that automates accountsreceivable and collections processes. In this role, Bussy isresponsible for the development of strategic products, services,and solutions. He joined Esker in 2002 as director of marketingcommunications, and in 2005 extended his responsibilities toinclude product management.