Robotic process automation — or RPA — bots don’t need a coffee break, they don’t get tired and they don’t lose focus after the 100th math problem that looks just like the 99 that came before. In other words, RPA is great for some of those peskier tasks finance and accounting teams don’t like to do.
For years, organizations have been trying to find financial improvements through enterprise systems, reporting tools and stopgap measures that attempted to eliminate repetitive manual actions.
“Unfortunately, what has been discovered is, while these solve many of the existing problems, they haven’t solved two of the core problems that exist in the finance and accounting world,” said Jeremy Dean, a national RPA lead at Mazars USA, a global accounting and advisory firm.
The first challenge was how to get data into these systems and the second was how to close their financials at month’s end, Dean said.
“Many are now finding that RPA provides the means for organizations to finally address and solve these problems,” he said.
How RPA works
RPA uses AI capabilities to reduce errors and execute repetitive, high-volume work. And it doesn’t need to automate an entire process to be useful. Instead, it can automate certain parts of a complicated process involving numerous steps, and that has been a major driver of its use.
“The quickest wins have been more rules-based processes that are more amenable to RPA,” said Dennis Gannon, vice president of advisory services at Gartner.
In this context, RPA could be thought of as digital duct tape.
“A good candidate [for RPA] is a task that is a bottleneck in a larger process that may take hours of manual work,” Gannon said.
For example, RPA is likely to be widely adopted as a means of automating tasks in the order-to-cash and procure-to-pay processes, he said. Starting with those processes allows finance teams to focus on the quick achievable RPA wins, get feedback on what works well, and then find more tasks that are easy to automate.
As to fears that the robots are coming for the finance teams’ jobs, it’s important to include those teams on RPA projects both to allay fears and to find new opportunities, Gannon said. Project leaders can start by inviting a few people from a finance team into an automation lab for a few days a month to practice putting new bots into a production environment. Over the course of the rest of the month they will notice how the bot worked and can identify any in-use problems or limitations. These deep dives can also teach them how to spot other automation opportunities between sprints in their daily work.
Dennis GannonVice president of advisory services, Gartner
“It’s not hard to see the opportunities everywhere once you get engaged,” Gannon said.
RPA consists of software robots, or bots, that represent a pattern of reusable automations for tasks and processes. Bots mimic some functions humans typically do, such as reading a screen in one application, copying the appropriate text, and then pasting it into another application. IT teams can use RPA platforms to create, monitor, manage, reuse and secure bots and their activities.
The simplest bots capture the rote human workflow and mimic it. This basic bot serves as a kind of template, which a bot developer can refine to create a stronger bot that is less likely to break if a screen on an app changes slightly. Bots are often complemented by other AI technologies such as optical character recognition (OCR) for capturing text from paper documents and machine learning to figure out which fields in an invoice map to fields in a finance application. When RPA is combined with other techniques, it is sometimes called intelligent process automation.
RPA can also complement other technologies for automation and integration. IT teams can sometimes use low-code/no-code platforms to create lightweight automations that are implemented as code. They can also use API management platforms or integration platform as a service to facilitate direct integrations that work much faster than RPA. However, RPA has an advantage in that it can access any application that a human can, which is not always possible or easy with these other technologies.
To understand how RPA is used in the real world, here’s a look at nine use cases for accounting and finance.
1. Automating governance
Scaling an RPA implementation can be difficult. One challenge is enabling finance departments to easily create new bots while also providing guardrails.
Hewlett Packard Enterprise (HPE) centralized its bot infrastructure to overcome these hurdles, said Sandeep Singh, finance global quality and RPA CoE lead at the multinational enterprise IT company, headquartered in San Jose, Calif. His team built out a bot platform on top of WorkFusion RPA. The bot platform helps simplify bot deployment and allows bot modeling, use tracking and error reporting. The team also created an internal governance framework to provide a complete view for stakeholders across audit, business compliance, IT and finance teams.
“This gives not just senior executives, but also the external auditors a sense of comfort that we have our eyes on the ball,” Singh said. “We constantly work towards ensuring that nothing slips through the gaps.”
2. Reconciling accounts
Comparing account balances between systems is a critical — but often tedious — function.
One of the biggest gains for HPE came from using RPA to improve journal entry and subsequent financial account reconciliations, Singh said. These processes are compliance-bound, time-consuming and involve disparate processes across the organization. For example, suborganizations within HPE have different templates, processes and approval flows. Some processes might involve functions within an ERP system. Some might involve audit and compliance requirements of identifiability for transactions, along with all the respective business requirements on approval flows and amount thresholds.
Singh’s team had to undertake a thoughtful process redesign to strike the right balance between the varied nature of business-specific requirements as well as stringent audit and compliance rules to create a streamlined process that his team could automate with RPA.
“While business requirements can be negotiable and are subject to improvisation, accounting rules and compliance requirements have to be dealt with kid gloves,” Singh said.
3. Processing invoices
Invoice processing is another promising RPA use case.
HPE’s accounts payable team processes a considerable volume of paper invoices each month and is responsible for recording vendor invoices for subsequent payment processing. Issues come in the form of multiple invoice formats, a range in the quality of scanned image invoices and the use of multiple languages. The company used a combination of OCR and machine learning modules from WorkFusion to mitigate image format and quality challenges.
OCR helps digitize the invoice images into a consistent text format. Machine learning then uses historical rules to determine how invoice fields map to finance applications. HPE uses RPA to post the extracted data into its SAP Ariba system for procurement management.
4. PO processing
Purchase orders often come in a variety of formats that vary widely depending on the source.
“RPA can automate and speed this process up, as well as reduce human errors,” Dean said.
As with invoice processing, OCR can help read paper documents, and machine learning can help map data from the documents into the system of record. For example, Dean worked on one project for a brewer that wanted to automate PO creation within their SAP implementation. An RPA bot received input in two different formats, validated the completeness of the data input, then set up the PO shopping cart in SAP and submitted a request for its approval.
5. Remediating discrepancies
Once bad data enters the enterprise data ecosystem, it can quickly spread to multiple systems and data repositories. This can result in significant downstream data cleaning and correction work.
Dean has worked on several projects to automate this process quickly and efficiently by scanning the data, finding issues and bringing them to a team member’s attention for review. Finally, once the correct data is identified, a bot can programmatically correct the data issue across all impacted systems.
For example, Dean worked on one project with a logistics company that used RPA to identify discrepancies between the ERP system and the company’s reporting tool. The bot evaluates the discrepancy and uses various rules to determine if the issue comes from an error with the source data or the reporting repository. The resulting decision is flagged for review and approval by a team member. Once the team member approves the change, the bot makes the change in the appropriate system. This increased the data quality across all systems involved.
6. Processing cash data
Cash application is a critical function in the accounts receivable process. Applying the right payments to the right accounts and invoices is a process that includes multiple ways to introduce errors.
The HPE cash application team processes a huge volume of payments from customers in over 50 countries. This process often starts with bank statements that need to be rendered in the appropriate format and copied into the accounts receivable application for a given department or group. RPA automates the process of reading the bank statements and copying data to the appropriate fields in the accounts receivable application.
HPE has faced challenges that include varying bank statement formats, multiple languages and missing information that compound the work of accounts receivables analysts, Singh said. In response, his team has developed an RPA workflow that uses fuzzy logic to improve data identification and machine learning to avoid repeating previous posting errors. This has drastically improved accuracy of cash application and substantially reduced processing time.
7. Ensuring vendor contract compliance
Making sure that suppliers are adhering to the agreed upon terms is critical to compliance, and this area is another promising use case for RPA.
As an example, HPE’s contract compliance team is using RPA to help automate many processes involved in ensuring adherence to vendor contracts.
“The activity is highly manual, cumbersome and fatiguing,” Singh said.
RPA bots can scan through contracts and purchase orders and use natural language processing to extract key information such as discounts, rebates and penalty clauses. The bots can then compare this information with information from HPE’s ERP systems on the actuals to identify the gaps and highlight discrepancies.
The use of RPA has significantly reduced the manual effort previously involved in the process, Singh said.
8. Reporting P&L
Profit and loss (P&L) reporting is another use case for RPA.
Many enterprises use RPA to automate P&L reporting, particularly in companies that need to provide daily reports to management, said Manish Chawla, associate director, business performance improvement practice at Protiviti, an IT consultancy.
RPA can greatly reduce the quantity of manual, repetitive and time-consuming tasks performed by finance experts so they can focus on more valuable activities, such as P&L reporting, Chawla said. Many firms cut processing time significantly and provide earlier access to reports with much higher accuracy.
9. Managing data across multiple systems
Finance teams often need to manage data residing in multiple systems. RPA can help improve data management across these systems, while also enforcing business rules associated with the movement of this data.
IT teams can build RPA finance automation to trigger on certain events in these systems, or bots can be run at specific time when it is necessary to complete a process, Dean said.
Dean implemented one system for a banking and insurance company that wanted to improve various processes involved in master data management and financial account maintenance. For example, they used RPA to automate three back-office processes related to seizure of financial assets for customers based on official legal requests made by executors. This made it easier to kick off one process that could access various databases and freeze relevant amounts across various accounts and two different core banking systems with minimal team member interaction. People could then focus on more judgement-oriented tasks such as reviewing and validating the data being updated.