In fairly short order, robots have begun taking over in the corporate world. Don’t be alarmed. This is nothing like the feared “singularity,” that prophesied (if dubious) moment when machines become smarter than humans and then, to prove it, commence wiping us out.
But robots are indeed infiltrating finance departments, some other functions, and operational areas in a number of industries. For the most part, robots are being deployed to automate repeatable, standardized, or logical tasks historically handled by people. In finance and accounting, think procure-to-pay, order-to-cash, and record-to-report processes. Enterprise-wide, the number of potential use cases seems almost limitless, although most of the potential is so far untapped.
Collectively, these technologies are referred to as “robotic process automation” (RPA), although they aren’t enabled by electromechanical machines that have arms and legs. They are virtual robots; in other words, software that is programmed to mimic the keystrokes humans make in completing a process. (At least, that’s a definition of what might be called “basic RPA”; many experts include more advanced technologies, discussed below, under a bigger RPA umbrella.)
What distinguishes the robots is that “they can work 24/7, they are very accurate, they do exactly what you tell them to do, and they don’t complain,” says David Wright, Deloitte’s finance robotic process lead in the United Kingdom. In other words, they don’t have human flaws. And with the use of RPA expanding exponentially, the potential workforce implications — depending on who you talk to, anything from reductions on a scale that could wreak havoc on the global economy to a broad transference of head count to more strategic, value-add positions — are a matter of some debate.
While RPA is largely based on technological capabilities that have been around for some time (like screen scraping, work flow, and rules engines), it didn’t begin penetrating companies in a significant way until recently. “This has really started to accelerate just in the past 18 months,” says Cliff Justice, principal, innovative and enterprise solutions, for KPMG. “But it’s already in at least the pilot stage in maybe half of the Fortune 500 companies.”
Virtually all RPA observers expect the trend to continue surging. RPA vendors, not surprisingly, agree. “We are at the very beginning of a major disruption in the accounting world that will be as big for the profession as Excel was,” says Steve Palomino, director of financial transformation at Redwood Software. “Robotics will become an industry standard.”
Although RPA is in its early stages, acknowledges Dinesh Venugopal, president for digital and strategic customers at Indian business process outsourcing (BPO) firm Mphasis, “customers with BPO-like processes that are coming up for re-bid are all asking about it.”
RPA software is not complex compared with other enterprise technologies. Nor is it particularly expensive, despite its relative newness. “In the past couple of years we’ve seen quite an uptake in companies wanting not big, transformative process reform, but very tactical reform to get a rapid ROI with a very low-risk solution,” says Craig Le Clair, a vice president and principal analyst at Forrester Research.
“I term it a non-invasive solution,” he continues. “You don’t have to do any data-integration projects or have a lot of cross-department meetings. [The technology] really is just about understanding the specific clicks that a human is making and substituting a software routine, a bot, to eliminate a lot of non-value-added steps.”
Similarly, Wright says he tells clients that “this isn’t rocket science. Seeing a demo is like being in a finance shared services center, looking over somebody’s shoulder and seeing keystrokes being made and transactions being executed.”
Midsize companies generally haven’t yet made much use of the technology, but they could, according to Mihir Shukla, CEO of RPA vendor Automation Anywhere. “It’s simple and cost-effective, and they will see the same ROI as an enterprise customer on a smaller scale,” he says.
How much ROI are we talking about? Enough that the practice of saving on costs by offshoring business processes may quickly become obsolete. An offshore, full-time employee averages about 35% of the cost of an onshore FTE, according to a Deloitte analysis. But a robot would typically be less than a third of the cost of the offshore FTE, or about 10% of the onshore employee’s cost.
Venugopal says Mphasis is quoting prices for RPA that will save customers from 25% to 40% on outsourced processes. But traditional BPO providers like Mphasis are in a tough spot, notes Le Clair. “Their [method] over the years has been to substitute labor for technology, because labor is what they have an abundance of,” the Forrester analyst says. “But while a number of partnerships have been announced between the BPO firms and the RPA product companies, more than half of the implementations are being done by companies on shore within their own data centers.”
So far, financial services has embraced RPA more than any other industry, followed by health care, utilities, and telecommunications — what Forrester calls the “hubs,” says Le Clair. The common denominators among those industries are a large volume of “swivel-chair work” (i.e., data entry and switching among various applications), the use of centralized shared services, and heavy regulation.
But the research firm is also seeing growth in what it calls “the edges”: manufacturing, supply chain, warehousing, and even oil. “We believe that in the next three to four years, the hubs and the edges will come together, disrupting the way modern enterprises work,” Le Clair says.
In financial services, think of all the processes required to, for example, approve and fund a home mortgage, from processing the application to the background check, credit check, property assessment, and approval. “Imagine how fast these processes have to happen and how accurate they must be for the new online mortgage companies that are processing applications within 24 hours,” says Shukla.
For finance and accounting at financial services and other types of companies, in addition to the processes mentioned above (order-to-cash, procure-to-pay, and record-to-report), RPA is being used for accounts payable, vendor statement reporting, and travel-and-expense processing, among other tasks, notes Le Clair.
Drilling down to more-specific uses, one common application of RPA for accountants is journal reconciliations, says Palomino of Redwood Software. Others include unapplied cash, where money in a bank account hasn’t yet been tracked to a specific customer payment; intercompany transactions, or those between company divisions or subsidiaries; and creating consolidated reports, which typically involves manual processes for pulling information together, and reformatting and distributing it.
But the tasks companies have deployed RPA for so far merely scratch the surface of its potential, according to Le Clair. “There’s greater usage of RPA, but I would say it has a low penetration today,” he says. “There are thousands and thousands of processes that could benefit.” Forrester plots technologies in terms of their maturity; there are five phases, and RPA is still in the first phase, called the creation phase.
It’s worth noting that you don’t need dedicated technologies to accomplish the various tasks that RPA addresses. “Similar to spreadsheets, robotic software tends to be both use-case-agnostic and application-agnostic,” says Wright. “You can use spreadsheets for financial transactions, monitoring the recruitment pipeline, or planning a wedding. In the same way, you can use [RPA] across numerous functions.”
As promising as basic RPA is, the most interesting business applications for robotic technologies increasingly will be more sophisticated. These applications will apply machine learning, artificial intelligence, cognitive computing, or some combination of them.
Among the most advanced RPA-related tools with cognitive capabilities currently on the market, and certainly the best known, is IBM’s Watson. It’s not an RPA program itself, but rather a platform on which smart RPA capabilities can be built. The big key is the technology’s ability to sort enormous amounts of data — including unstructured data such as emails, which are typically text-heavy. By some estimates, unstructured data, which historically resisted deep searching and analysis, accounts for as much as 80% of all data in the world.
“If you’ve built bots on Watson, now you’ll be able to infer, reason, and interpret natural language, and the context of that language, and really automate more decisions that in non-intelligent bots would get escalated to humans,” says KPMG’s Justice. “Watson will gather the evidence available and make a probabilistic determination.” Notably, procurement organizations are using Watson to help search for alternative, lower-cost vendors.
Such technology can greatly improve invoice processing, for example. Often, an invoice must be read for context. It might have data in the wrong field, or lack information needed to determine whether it meets contract requirements or procurement rules. It might require a three-way match, verifying with the business, the vendor, and the procurement organization that the correct products or services were delivered. A cognitive technology based on machine learning and AI can do all that.
And cognitive bots get smarter over time. They can monitor how humans interact with the system, and as more data on those actions is collected, the machine gains confidence that it can process exceptions or handle other issues, notes Justice. Credit card fraud detection is a perfect example of machine learning that’s in common use, with bots getting smarter and smarter at detecting when cards are being used suspiciously, says Venugopal.
Three years ago, companies’ online chat bots weren’t very impressive. “Now they’re being built on programs like Amazon’s Alexa, which is still very new but has a deep, machine learning capability,” Justice says.
Call centers are also perfect candidates for improvements based on machine learning. A company starts by encoding standard operating procedures, and then today’s technologies track voice interaction with customers and convert it to machine language.
Also, the platforms behind personal digital assistants Siri (Apple) and Cortana (Microsoft) have been open-sourced, allowing the development of advanced RPA capabilities based on speech-recognition technology.
Advanced RPA could even play a major role in closing the books. According to a Forrester report, at financial services firm UBS, 2,000 people are involved in the closing process. The company told Forrester that it believes it can fully automate the process within five years.
These advancements are not as much of a technological miracle as they may seem. The key is developing the curated knowledge base that such tools will draw from. “The algorithms are fairly well known, and they’re going to become commoditized,” says Le Clair. “What you need are domain experts to develop the right model for how humans interact with the machine so it can take advantage of human knowledge. We’re just at the point where we’re starting to see the knowledge bases and domain expertise needed to take this to the next level of being very real and practical.”
At the same time, there is a danger, according to a Forrester report released in June, that company leaders will get too caught up in the automation whirlwind and overautomate, thereby degrading the customer experience. “Those who think that ‘everyone is automating everything, so we should drive as much cost out of our business as possible’ will drive customers away,” the report said.
“A common denominator of RPA approaches is decoupling routine service delivery from labor arbitrage,” says Tom Reuner, research vice president, intelligent automation, at HfS Research. Translation: in an RPA-centric business environment, companies won’t have to think anymore about whether to offshore back-office processes to take advantage of cheaper labor. That’s because labor needs will be drastically reduced, and running a bot will cost the same — that is, not very much — everywhere.
And it’s not only offshore jobs that will go away. Forrester estimates that RPA and machine learning will cause the number of U.S. “cubicle workers” to decrease by 16%, or 12 million workers, by 2025. KPMG suggests the worldwide total could be as much as 100 million jobs.
However, Forrester said, the 16% of U.S. workers displaced will be partially offset by the creation of new jobs as a result of these technologies, equivalent to 9% of the current total cubicle jobs, for a net decrease of 7%. “Not transformative, but still a fair number,” says Le Clair.
But Wright suggests that there are alternatives to the “now we can cut heads” point of view. “Rather than remove people, some organizations will keep the headcount they have but take in a lot of extra work,” he says. For example, he told of a company that was providing a certain reporting service to one unit in the business. When robots came along, the company decided it would now be cost effective to provide the service to the entire business.
Shukla of Automation Anywhere offered an even rosier outlook for back-office workers. “The workforce will change and skills requirements will shift,” he says. “Our customers tell us they’re seeing sighs of relief from employees, who hate the mundane tasks they have had to do over and over. Imagine human employees metaphorically working next to software bots. When the humans have something that must be done repeatedly, they hand it off to a bot.”
A new digital workforce that makes the finance department more productive and handles tasks staffers would rather not? It sounds like an idea most finance departments would heartily embrace.
There may be as many as 2,000 startup companies trying to marry robotic process automation with machine learning and artificial intelligence in a bid to fundamentally change the way we work.
But currently, a much smaller cadre of companies is dominating the field, including those providing basic RPA and those developing more advanced approaches. The lines between those categories are significantly blurred, but consider the following players to be worth investigating:
|• Arago||• Nice Systems|
|• Automation Anywhere||• OpenSpan/Pegasystems|
|• Blue Prism||• Tata Consultancy|
|• IBM||• UiPath|
|• IPsoft||• Wipro|
|• Kryon Systems||• WorkFusion|