Applications of AI in Accounting
According to a survey by the World Economic Forum, most executives from 151 financial institutions in 33 countries expect that Al will become essential to their business within two years (World Economic Forum, 2020). Although Al receives much attention in corporate news, there is a large gap between business leaders who believe that Al can provide their companies with a competitive advantage versus those who are already using it. In a global survey of over 3,000 executives by the MIT Sloan Management Review (in collaboration with the Boston Consulting Group), nearly 85% of respondents reported that they expect Al-enabled technology to provide a competitive advantage for their firms. However, less than 39% reported having an Al strategy in place, and only 5% extensively reported having incorporated Al into their products and services (Ransbotham et al., 2017). This study exemplifies the current environment for Al: business leaders recognize the potential benefits that Al can bring, but widespread adoption has yet to be achieved. The opportunity for accountants to leverage Al for business is tremendous.
The power of present-day Al technology far exceeds accountants’ traditional technological toolkits (e.g., electronic spreadsheets, general ledger software, and tax compliance programs), especially when combined with advanced data analytics. Artificial intelligence has three broad applications for business: (1) to better understand and interact with customers, (2) to offer more intelligent products and services, and (3) to improve and automate business processes (Marr & Ward, 2019). Because accounting is a highly process-oriented discipline, many of the current use cases automate traditional accounting functions. Accountants who understand how Al is currently being used in practice are well-positioned to suggest other possible Al-enabled solutions to their unique business challenges. To this end, this chapter provides examples of how Al can be deployed across various sectors of the accounting industry, including financial accounting, management accounting, audit, tax, and advisory services.
Financial Accounting Applications
Cash and Account Reconciliations
Accountants understand that performing bank reconciliation serves as a critical detective control to help safeguard cash and improve the accuracy of reported accounting information. However, completing traditional bank reconciliation can be time-consuming, especially for mid-size to large companies with multiple accounts at various institutions. Common errors encountered during the bank reconciliation process include duplicate entries, different data formats between systems (e.g., bank portal versus ERP system), and human-generated data entry errors (e.g., misspellings, character spacing) (Sigma IQ, 2019). Automating the bank reconciliation process with artificial intelligence promises to save time and improve accuracy for accounting departments.
Expert Systems (ES) are well suited for bank reconciliations because the rules are straightforward and do not change over time (Siegel et al., 2003). Nolan Business Solutions, an international firm with headquarters in the UK, uses an ES called Advanced Bank Reconciliation (ABR) that works with NetSuite, a leading provider of cloud-based services including accounting and ERP systems. ABR imports transactions from the bank, then automatically matches transactions based on user-defined rules, and finally generates reports on un-reconciled transactions (Nolan Business Solutions, n.d.). The ABR interface allows users to classify any remaining unmatched transactions, significantly reducing the amount of time that an employee would spend manually matching the transactions. After the reconciliation is completed, the transactions are saved to the hard drive or network. The transactions can be viewed or examined later for audit purposes.
The latest Al technologies, such as machine learning (ML), can further enhance the automation of account reconciliation. For example, Sigma IQ is a software provider that specializes in using machine learning to automate the account reconciliation process. Its cloud-based Al platform goes beyond a rule-based Expert System by deploying algorithms to develop a nuanced understanding of how to resolve common issues. In this way, ML can continuously learn each time that a human makes a mistake and include that error in its
Applications of Al in Accounting 21 processing for future reconciliations. As such. ML can significantly improve the account reconciliation process and save companies hundreds of manual hours in processing time (Sigma IQ, 2019). ML can also identify and prioritize reconciling items that might need further investigation by staff accountants.
Receivables and Sales
The cash application process requires accounts receivable staff to match incoming payments to outstanding invoices. This timeconsuming process is costly and prone to errors, especially when dealing with many customers and high volume transactions. This process has become even more complicated due to the variety of payment methods available (e.g., check, ACH, wire transfers, credit card payments, PayPal, Venmo, etc.). Additional problems can result from missing or incorrect reference numbers, combined or partial payments, and invoices in different languages and currencies (Sinha & Davis, 2018).
Al has helped firms to automate processes that match payments and remittances at much faster speeds than can be done manually, effectively reducing the number of staff hours and costs associated with this activity. ES can be used to validate customer orders, grant credit based on pre-defined criteria, automatically invoice, and post-sales entries to subsidiary ledgers (Siegel et al., 2003).
More recent developments in Al and ML technologies are providing additional efficiencies and cost reductions. For example, Fifth Third Bank offers a service, called Expert AR Receivables Matching, which integrates with ERP systems to reduce the cost and risk of manually matching and posting receivables (Fifth Third Bank, n.d.). The service works by collecting all payments into a centralized location and then using ML to match and post A/R transactions. Once a payment exception from a customer is manually validated, the system learns to recognize incoming payments in the same format to increase the chance that it is processed without exceptions. Unmatched payments can be reconciled with a single click using a matching algorithm.
Citi is another bank leading the industry in deploying Al to manage their receivables. They developed a solution called Citi® Smart Match. Smart Match uses Al to read a variety of sources of remittance information (e.g., emails, email attachments, faxes, remittance advice, web portals, and electronic data interchanges (EDI)). The purpose is to identify and extract the most important payment details (Sinha & Davis, 2018). The technology then transforms the extracted data into a format that canbe used for creating structured remittance data files. The remittance file is then matched against the company’s outstanding receivables, and the results are transmitted to the company’s ERP system “on a straight-through basis to achieve end-to-end reconciliation” (Sinha & Davis, 2018, p. 2).
The Citi® Smart Match system identifies unmatched items and generates a report in much less time than it would take a human to complete. A/R specialists can then use the report to manually input missing information or correct erroneous data. ML is used to learn from each human interaction and resolve future unmatched items on its own. It typically takes three to four months for the system to be trained, and then it can achieve straight-through reconciliation rates near 90%, reducing the amount of time spent by staff updating errors by 80% (Sinha & Davis, 2018). Thus, the amount of errors decreases, operational efficiency increases, and payments are posted faster, freeing up A/R staff to focus on the collection of valid cash items. Performance metrics such as Days Sales Outstanding (DSO) decreases, and the working capital improves.
Finally, another example is HighRadius Corporation, a provider of Al software for accounts receivable operations, including A/R automation, predictive capabilities (cash forecasting), and analytics. HighRadius asserts that its cloud-based platform is used by more than 200 Fortune 1,000 Companies, including global firms such as Starbucks, 3M, and Johnson and Johnson (HighRadius, n.d.). One of the case studies featured by HighRadius includes Express Employment Professionals, a staffing and recruiting company in North America. Before using HighRadius, Express Employment faced several challenges related to its A/R operations, including high fees for lockbox services at three banks averaging over $7,000 per month in total. The company processed 25,000 invoices per week, 4,500 line-items per day, and 3,000 payments per day. Furthermore, the ACH payment process was 100% manual, and payments were de-coupled from remittance sources such as emails, portals, and EDI files. By using HighRadius’s Al platform, Express Employment achieved 85% automation for checks and ACH transactions and saved $84,000 in annual lockbox fees (Richards, n.d.).
