By Ted Howard
Robots are a good solution for many processes, but there is a cost and without a thorough review of where it’s applicable, that cost could be high.
Robotic process automation (RPA) is a hot topic in many treasuries and other finance functions in companies globally. The promise is that companies can eliminate the risk of human error on tasks that are repetitive and rules-based. This is for the most part true. But does the return on investment live up to the promise?
That’s the question that arose during a recent NeuGroup Internal Auditors’ Peer Group meeting. One member who presented acknowledged that RPA reduces the need for repetitive human effort, but said that there were distinct limitations to the types of work where it can be applied. Further, human effort is likely needed at the front of the process; that is, RPA can’t structure the data to be digested by the computer program.
The attraction of RPA is that the technology is more readily available, effective, and affordable than its higher-level cousin, artificial intelligence (AI). This means RPA can be a good low-tech starting point to replace repeatable processing that is traditionally performed by a human; tasks that are routine and simple but high volume. In several pre-meeting surveys across the NeuGroup universe, many members have said they are in some stage of RPA. Many in a more exploratory way and others more operationally.
But companies looking to embark on an RPA project should take a look at the tech feasibility as well as the ROI. “RPA is so hot that people want to implement it without thinking it through,” said one member of the IAPG. He added that if RPA was “just coming into being perhaps five years ago, I could see this really catching on,” he said. But right now, companies have solved for the cost issue by outsourcing to lower-cost countries. “Highly educated, low cost workers in some countries can still get these mundane financial tasks done quickly and more cheaply,” said another member of the group.
In addition, current RPA, according to some members, still requires the human touch. “We always have manual intervention in RPA software,” said one member. Added another, “There are other steps outside of the automated processes that still have to be done by a human for RPA software, so what is the real value?”
“RPA has potential and ROI can be realized in less than a year’s time,” said Sharad Chandra, head of RPA at Aibots Technologies LLP. But in its current phase, too many companies are not taking the time to do detailed analysis of where RPA fits; or how it works, he said. Thus the realized benefits have been underwhelming.
Further, the market isn’t hearing about the failures because “customers don’t want the world to know about it,” Mr. Chandra said. “Sometimes when you try and fail at something you don’t want the world to know about it.” He cites several instances where the results have disappointed: a major bank that is three years into an RPA project and seeing a maximum ROI of 7.2%; or a major “RPA player” that claimed to have developed 9,000 bots but “none of them were finished within budget or schedule and only 4-5 percent delivered ‘some’ results.”
On top of the complications, RPA is going through somewhat of a “fad” stage, thus, the cost is “bloated,” Mr. Chandra said. But prices should come down as more people and more companies get more familiar with the concept.
With this in mind, experts suggest pinpointing exactly which processes are the best candidates for RPA and also understand that robotics should be seen as just one component of an end-to-end process improvement—not the end-all be-all solution. Also, start with very simple tasks. Deloitte, which recently presented on RPA before NeuGroup’s Asia CFOs’ Peer Group (ACFO), recommends testing RPA on a process that is small scale to gain familiarity and to score an easy win. After that, the consultant said, progress to increasingly large and complex environments, funding subsequent stages with savings from prior stages.
According to Deloitte, RPA can be used at many points along the “finance value chain,” as it calls it, but there are levels of risks to each function. For instance, for accounts payable, RPA has a higher level of receptiveness for things like perform vendor reconciliations (risky), prepare payment runs (modest risk), and release invoice (low risk). But it cannot manage queries or approve invoices. For accounts receivable, RPA will have less impact, with processing sales orders (low risk) about its only application. Generally speaking, RPA is better suited to AP, AR, fixed assets and T&E, versus cash management and general accounting.
According to Mr. Chandra, companies embarking on an RPA project must “understand the features and limitations of RPA as a concept” and treat it as a concept/method rather than as a technology.
In the final analysis, the costs and effectiveness will, like many things, depend on the company. “I think the financial ROI depends on specific company situations,” said one IAPG member. That is, “where their functions are located and whether they have already streamlined the business processes.”