Robotic process automation: A cheat sheet

Here's what you need to know about robotic process automation, including major software vendors, how it might affect jobs, and how to start using RPA.

RPA, Robotic Process Automation Concept

Image: Olivier Le Moal, Getty Images/iStockphoto

One of the most noticeable effects of work in the digital era has been the rise of tedium. Many simple tasks require a bevy of clicks, cuts, copies, and pastes that eat up employee time on tedious tasks, yet remain completely necessary. Fortunately, there's a business automation tool that can, in theory, eliminate all of that tedium: Robotic process automation (RPA). 

RPA has the potential to shift many mundane tasks to software robots and AI workers who can take care of busy work and free up employees to do more important tasks. RPA does have its shortcomings, but businesses whose employees spend a lot of time entering data, generating invoices, and performing other simple workflow tasks may want to consider an RPA solution. 

SEE: Robotics in the enterprise (ZDNet/TechRepublic special feature) | Download the free PDF version (TechRepublic)

What is robotic process automation?

Robotic process automation does exactly what its name implies: It uses robots to automate basic workflows, saving humans the time it would normally take to do predictable, repetitive tasks on a computer.

At its most basic level, RPA isn't any different than any other form of automation, which generally aims to figure out which human tasks could be done by a robot and then designs the robot needed to do the task. RPA approaches automation from a different angle: The graphic user interface (GUI) of modern desktop software.

Instead of building a robot designed to perform an action, which can require specialized hardware and deep layers of machine learning, RPA builds software robots designed to follow workflows the same way a human does: By clicking, cutting, copying, and pasting data between apps, filling out digital forms, sending emails, and interacting with computers at the user level. 

That means RPA software bots navigate computer programs the same way a person does, and they're trained the same way, too. RPA bots learn by recording macros from watching humans use GUIs to perform repetitive tasks and can then be run to perform those same tasks without human intervention or monitoring. 

SURVEY: What effect will the use of robotic systems have on jobs in your industry?

RPA software bots don't need computers attached to monitors, either--as long as the system they're running on has access to the appropriate files and has the appropriate apps installed they can, in theory, keep performing the task ad infinitum.

One of the biggest advantages to robotic process automation is its simplicity. With only a bit of setup from IT, accountants, data processors, administrators, and others who perform routine tasks on a computer can create their own RPA software bot without the need to dig into code to do so. In an ideal situation, non-IT workers would be empowered by RPA tools that they could train and hand off work to, creating a new class of "digital employee" that could eliminate the mind-numbing tasks that humans hate to perform.

Anyone in the IT or software development world who has used an automated GUI testing tool is probably sensing some familiarities, and rightly so. Software like TestComplete, IBM's Rational Function Tester, and other products that automate the testing of GUIs function in similar ways, albeit with a broader scope and purpose: Instead of doing the same single task over and over again, GUI test tools are designed to dig through the entirety of an app's interface to look for performance problems. 

Additional resources

What are potential uses of robotic process automation?

There are a lot of ways in which RPA can be used to offload repetitive processes, and it's difficult to completely cover all possible use cases. Each business should ask itself, and potentially survey employees, to get a good feeling of the sorts of tasks that different departments would feel comfortable trusting to a robot.

Depending on the type of task to be automated, one of two types of RPA bots can be used: 

  • Attended: Requires a human to trigger, and generally automate part of a task while leaving the rest to a person. 

  • Unassisted: Acts automatically when a certain condition is met, typically by the bot monitoring a queue for new jobs. Unassisted RPA is also known as RPA 2.0 and uses elements of AI and machine learning to increase its independence.

There are cases when either, or both, types of RPA software robots can be used, such as:

  • In customer service, where attended and unattended bots can smooth the service process and make face-to-face and digital communications easier for employees and customers;

  • In health care, where lots of different types of data is tracked, manually input, and then has to be matched up to other databases, kept up to date, and otherwise eating up a lot of time for nurses, doctors, and medical coding teams;

  • For generating invoices, an RPA bot could take new customer information input into a spreadsheet and generate quotes or invoices that are then automatically emailed to accounting and the customer;

  • Digital transformation initiatives, as well as migrating from one software platform to another, often involve a lot of transcription and copying of data. RPA bots, when trained properly, could do a good portion of that work.

  • IT teams can be greatly sped up by RPA bots that can preemptively spot problems to eliminate help desk calls, pull system information for techs, and do many of the basic system actions that helpdesk professionals have to perform, like rebooting machines, installing software, etc.; and

  • Warehouses can speed up inventory management by automating inputs and reports.

This list could go on forever--B2B AI solutions company AIMultiple lists 61 separate use cases for RPA, and there's likely even more than those.

Additional resources:

Is robotic process automation practical for today's businesses?

Here's where the real sticking point of robotic process automation shows up: It's hard for RPA to be scaled up to work for large organizations. 

Deloitte, in its 2020 global RPA survey, finds that RPA adoption is growing fast, but only 3% of organizations using RPA have managed to scale their use beyond 50 bots. Deloitte said that indicates that companies using RPA are having more difficulties than anticipated.

Deloitte expressed surprise with finding RPA scaling to be so low, especially "given the relatively low cost of implementation and the high benefit of automating carefully selected high-value activities." The report does admit that RPA is still in its infancy, and many companies exploring it may still not have goals aligned to properly use RPA to its fullest extent.

Unfortunately, one of the major difficulties in rolling out RPA beyond a few instances is inherent in its design--it relies on using software GUIs to function. If an update is rolled out that changes the location of a single menu item, a button, or a field, then the RPA bot using it breaks. 

SEE: All of TechRepublic's cheat sheets and smart person's guides

Writing for Forbes, analyst Jason Bloomberg warns that businesses should think twice before implementing RPA due to the fragile nature of RPA bots that can stop working for very slight reasons. The problems, he points out, are even worse when dealing with legacy software that lacks APIs. With an API, an RPA bot can be configured to learn from interface changes and work with the app's back end to overcome GUI limitations. Eliminate the API, and you're stuck with relying completely on an app's GUI to remain unchanged.

Bloomberg's sentiment is echoed by Sanjay Srivastava, chief digital officer at digital transformation strategy firm Genpact. Most RPA implementations fail, Srivastava said, because companies don't think of RPA as one part of a larger strategy, instead looking at it like a fire-and-forget tool, which it isn't.

"Many companies think that once bots are set up, they will just run in the background and operate autonomously. The reality is, bots need constant management and maintenance over their productive life," Srivastava said. 

Maintenance doesn't only include updating RPA bots to account for software GUI changes, it also means being sure bots are configured to automatically know when passwords to apps and directories change, being aware of changes in the structure of data a bot needs to perform its job, and that bots are updated for process changes. 

So, while robotic process automation can theoretically do a lot for modern businesses, it's important to realize that it's a relatively young technology that requires constant upkeep and integration with larger digital strategies, which very few organizations have been able to do successfully so far.

Additional resources

How will robotic process automation affect jobs?

There can't be a discussion about any kind of automation without talking about its effect on jobs, and robotic process automation is no different. 

With RPA, the big question on the minds of people employed doing repetitive tasks is whether their jobs are at risk. The answer to that question, said Villanova University business tech professor Steve Andriole is: Yes, absolutely and positively.

"If you perform routine tasks or even what appear to be complex deductive inferential tasks we associate with 'knowledge' industries, yes, your job and career are at significant risk: AI and RPA will absolutely, positively threaten your job, your career, and your very professional existence," Andriole writes in Forbes.

Andriole likens low-level AI and automation like RPA to the invention of the automobile, and the humans doing RPA-vulnerable jobs to the horses and coach-builders that cars displaced. Stating otherwise, he said, would be "irresponsible and naive."

SEE: Managing AI and ML in the enterprise 2020: Tech leaders increase project development and implementation (TechRepublic Premium)

Others argue the opposite, which has been a common refrain for proponents of automation and RPA: It will free up employees to do more valuable tasks, reduce stress, and make each employee more valuable, said RPA provider UiPath.

To get a better idea of whether or not jobs are actually at risk, it may be helpful to turn to a report on RPA from accounting firm KPMG. Look at the language used in the report and you'll get a better idea what financially-concerned organizations see RPA as: 

  • "RPA may be the final nail in the coffin for labor arbitrage."

  • "It's been estimated that a software robot can cost around a tenth of a full-time worker in the United States, United Kingdom, or Australia."

  • "Success in today's complex global financial markets requires unprecedented levels of speed, accuracy, and cost efficiency beyond what a human workforce can provide."

  • "We believe that the industry as a whole must move quickly towards adopting RPA and AI, especially in light of the increasingly complex regulatory intensive environment,"

  • "Machines can do many tasks better, cheaper, and faster." 

  • "[New Zealand Banking Group Limited] uses bots in finance, human resources, payments, and mortgage processing departments as well. It was noted that in the payments area the number of human employees has decreased from 40 to two."

The KPMG report further states that "some fear that smart robots may replace more than 100 million knowledge workers—or one-third of the world's jobs—by 2026." It also adds that employees have "good cause for concern," citing massive Wall Street job losses that "will continue throughout the financial services market." 

The report also details steps to get worried workers onboard with RPA, telling them that the goal of RPA isn't to reduce headcount, before immediately pivoting to say "head count may end up being reduced." 

So, while it may be hard to say with absolute certainty whether RPA will replace jobs, those who have studied it, like Andriole, and those who stand to benefit from reduced labor costs both seem to be giving an emphatic yes.

Additional resources

Who are the major robotic process automation software vendors?

There are a lot of companies built around offering robotic process automation software--a quick Google search will return dozens of them. That said, there are a few that seem to be the leaders in the industry. 

UIPath, Blue Prism, Automation Anywhere, and Kryon Systems seem to be continually at the top of the pack in terms of customer ratings and name recognition. 

Microsoft also offers its own RPA platform, Microsoft Power Automate (previously called Microsoft Flow), so businesses with a strong investment in Power BI, Azure, or other Microsoft products may want to consider its offering for easier integration. 

All of the various RPA companies offer similar services, so the vendor an organization ultimately chooses may be due to one or two differences that suit their particular needs. If you're looking for an RPA vendor, it's a good idea to shop around, talk to customer service, and find out which one is right for you. There are a lot of factors that go into picking an RPA vendor—far too many to cover here.

Additional resources

How do you get started with robotic process automation?

Turning your organization into one that leverages robotic process automation is a huge undertaking. Like any business transformation initiative, it takes careful planning to do it right.

TechRepublic's cheat sheet on digital transformation offers several ways to approach an RPA project. There are also RPA initiative tips to be gleaned elsewhere on TechRepublic and at its sister site ZDNet.

That said, there are some essential things to consider when getting started with robotic process automation, or any other form of digital transformation.

1. Figure out where you currently stand

Before you can plan an effective robotic process automation implementation, you need to know what you hope to achieve and you can't know that until you know where you currently stand and what processes you want to automate.

Talk to some of the departments that are good RPA candidates, like finance, sales, and other groups that do a lot of data input. See what they think would be good applications of RPA, and then do some modeling to see if it would actually be worth the time and investment for a particular use.

2. Formulate a game plan

Once you know what you want to use RPA for, it's time to think about how to achieve those goals. 

This is the time to start talking to vendors to find out which is the best fit, building an RPA team, and training IT on the tasks they'll need to take on to make RPA successful.

This is also the time to get those who will be using RPA day-to-day involved in creating workflows for bots to learn and use. If you can go into the execution phase with well-outlined plans down to the smallest detail getting started will be much easier.

3. Test, implement, and expand

Once you have plans in place, have chosen a vendor, and have an RPA team assembled you can work on the four phases of implementing RPA that RPA vendor UiPath illustrates on its website:

  1. Build a proof of concept.
  2. Run a pilot program in a production environment.
  3. Ramp up performance based on feedback from the pilot program.
  4. Integrate RPA and make it part of your institution.

There's a lot of work to be done to build a robotic process automation system from the ground up, and these phases could take months, even years. Like any transformation initiative, there will be hurdles, and overcoming them will be key to short- and long-term success.

Additional resources