As the forthcoming book, The Knowledge Work Factory, argues, there is only one underlying element common to all business improvement, and that is standardization. Specifically, this can be traced to the world of manufacturing, and its three timeless principles of standardization, specialization, and division of labor, famously described by the economist Adam Smith back in 1776. Yet as knowledge workers have come to represent the bulk of today’s labor force, their productivity gains have not kept pace with modern computing power; this is the infamous Productivity Paradox first described by Nobel laureate Robert Solow in 1978.
Yet knowledge workers also lag far behind their manufacturing counterparts as well, in terms of productivity gains over the last 100 years. Now, however, it appears that robotic process automation, or RPA, is poised to come to the rescue. In this article, we will explore the roots, realities, and possibilities of RPA, and how you can apply them to your business. We’ll cover its history, things to watch out for, how to implement it, and the different types of training available.
The past 30 years have witnessed a multitude of new core IT systems being implemented in businesses. These include on-premise systems, cloud-based applications, and a wide range of desktop applications, all being used daily—and simultaneously. The resulting landscape, strewn with disparate systems, has left a long tail of micro-level data-processing gaps between them all. These system-to-system gaps are not being filled by other systems; currently, they’re being filled by people. In other words, there’s a lot of copying-and-pasting going on today to connect every System A with System B, not to mention C, D, and E, etc. It’s clearly a ripe opportunity for the convenience and processing power of robots.
Using robots to process work is not a new concept. As most people know, robots have been used on assembly lines for decades to assist in the production of automobiles, as well as other manufactured products. With RPA, white-collar, or knowledge-work processes, now have the opportunity to benefit from the equivalent of factory robots—with a speed of implementation and processing that promises a seismic shift in knowledge-work operations.
Just looking at search-history trends, you can see that robotic process automation interest in growing.
Robotic process automation is defined as the implementation of software robots that are configured at the graphic-user-interface level on individual computers, end-user devices, or by work-station—and in turn used to automate routine and repeatable tasks currently conducted by white-collar or knowledge-work employees. The robots themselves are created in applications such as UiPath, Blue Prism, or Automation Anywhere.
RPA was created to enable automation at the graphic-user-interface or GUI level, allowing users to see what they’re creating. This distinction is essential: It allows employees who lack computer-coding skills to connect disparate applications all by themselves. Prior to the advent of RPA, this required considerable back-end system-level coding by trained IT professionals, using application programming interfaces or APIs or scripting. Thus RPA can be seen as the democratization of programming, handed directly to the knowledge workers who need it most.
Robotic process automation is a fancy name used to describe software robots that automate manual routine data-entry and transcribing tasks across different systems which burden knowledge workers—and consume up to 50 percent of their days.
Robotic process automation seeks to solve the “long-tail problem” of automation and the overwhelming number of micro-level tasks required to fill data-processing gaps between core IT systems and disparate desktop applications that exist—and persist—throughout knowledge-work business operations today.
In theory, robotic process automation can be used to automate entire processes. But in more common and achievable applications, RPA is used to act as a “micro task” bridge automation tool that closes data transcribing gaps between core systems and data, email, or word-processing applications. Think of it as a macro that can switch between apps.
The below screen shots show how RPA can scrape data from closed PDF files into an Excel document with ease.
With robotic process automation, all types of industries can speed up the processing of transactions, streamline operations, automate responses, transfer data, and bridge communication with other systems.
Common work tasks that can be automated via robotic process automation include:
These are all processes which are currently performed manually, making them tedious, tiresome, and error-prone.
In the finance and accounting world, for example, RPA can mean your workers spend less time on invoicing-processing-related data entry, gathering, and transcribing, allowing them to spend more time on decision-making functions, like analytical review and presentation creation.
As another example: in banking, RPA lets workers integrate “the last mile” across business units and the back office themselves, enabling banks to automate data related retail branch processes, commercial and consumer lending administration, loan processing reconciliation, underwriting, and anti-money-laundering.
RPA also helps with the “last mile” in health insurance claims. RPA insurance use cases involve automating simple, repetitive tasks that take up administrative and clerical time, such as copying and pasting batch data from an email into an underwriting application. By automating these routine tasks, health insurance claims staff can focus on more important, high-level work.
1. Do you need to know coding for RPA?
No, but knowing basic code is a plus.
2. How much does robotic process automation cost?
The cost of the UiPath Commercial Products license depends on the number of robots and developer studios chosen. The average cost is roughly $100K for medium scale implementations – and we are talking about software only.
Once again, that is just software cost. You can easily add 200-300% of consulting cost on top if you need use cases identified and robots configured too.
3. How long does installation take?
This depends on the length and complexity of the process.
Basic robot configuration will take about 4 days to build and test.
Other, more complex “linked” bots have taken as long as 2 weeks.
4. Is this something an analyst can learn?
Yes, although becoming skilled can take 2-6 months, depending on the individual and the time dedicated to training.
5. Can a line of business run their own projects (instead of a centralized IT team)?
A dedicated analyst can build, test & manage bots from their own UiPath studio.
If other lines of business have similar business needs, the UiPath Orchestrator can deploy
processes to other bots.
Creating a “Robotics Center of Excellence” is the easiest way to ensure RPA implementation standards across the company, govern robots, and ensure RPA success.
You don’t need to be the CIO or CTO of a major corporation to enjoy near-term benefits from implementing RPA software in your company’s processes. Today’s business-unit leaders, mid-level managers, and front-line employees are taking on robot creation on their own—without the help of their IT departments, in many cases.
It’s easy to see why. Simply consider the top 12 benefits of RPA that any businesses installing, or considering implementing, robotic process automation in their business can expect:
Six real-life examples of robotic process automation case studies in Fortune 500 companies
The history of RPA closely resembles the evolution of its mechanical predecessors: industrial robots. The first true industrial robot was created in 1954 by robotics genius George Devol (owner of Unimation Corp.) to automate the moving of items from one place to another less than 12 feet away. Just a few years later, in 1961, General Motors began adopting these robots on a large scale across its assembly lines.
The first industrial robots were simple. They were used to automate tasks such as picking up, moving, and placing items on the assembly line. Yet as new technological breakthroughs appeared—such as sensors and cameras which allowed the robots to “feel” or “see” what had to happen next—they quickly grew in both complexity and capability.
Fast-forward to 2018: There are now nearly two million industrial robots being used in factories. And the pace of robots being installed is estimated to rise 14 percent annually through 2020—meaning that almost three million robots will working in factories by then.
The principles of robotic process automation mimic those of their industrial predecessors. Whereas industrial robots moved things from one place to another, RPA was created to move data from one place to another. It’s evolving in a similar way to industrial robotics, suggesting a similar trajectory.
Is RPA an invention unto itself? Or is it merely an innovation based on previously-existing technologies? These questions spark debate. For example, early software bots have been automating social-media posts, sending emails, and responding to basic web chat since the internet started taking off in the early 2000s. Similar software—in the form of interactive voice response or IVR—has also been used in call centers for over a decade. Still, neither of these technologies accurately represents RPA’s capabilities, nor what it empowers businesses to accomplish.
Actually, a significant portion of RPA’s lineage can be traced back to early screen-scraping tools and management software. Understanding these will help you better understand—and appreciate—RPA:
While each of these technologies represents an essential component of robotic process automation, RPA’s whole is greater than the sum of its parts. RPA takes the most useful pieces of each of these recent technologies and refines them into unifying, cohesive software that’s easier for the average user to control (especially when compared to writing code!), and far simpler for practically any business to implement.
RPA is radically different than the solutions that came before it because it’s not dependent upon a specific programming language or a particular application. Instead, it runs at the display, or surface level, of the process. What this means, in practical terms, is that issuing commands, managing workflows, and integrating new applications can be accomplished with drag-and-drop simplicity.
Furthermore, RPA takes advantage of optical character recognition (OCR) technology which allows it to adapt to changing websites without slowing down or requiring employee intervention. Today, we can think of robotic process automation as a configurable AI, perfect for optimizing business workflow.
While the overall history leading up to robotic process automation software may be long, its most explosive growth and development have occurred over just the last two years. During this time, RPA has become one of the fastest-growing industries in business and technology.
On March 18, 2016, Blue Prism made history by becoming the first publicly-traded RPA vendor. Then In 2017, the innovative and popular RPA startup UiPath made news when it received over $30 million in Series A funding that allowed it to greatly grow its staff and keep up with the rapidly expanding demand for automation bots. More recently, Automation Anywhere announced in June, 2018 that it had acquired over $250 million in additional funding, with CFO Clyde Hosein stating, “with this investment, we are poised to extend our leadership in the multibillion-dollar RPA market.”
When considering the fast-approaching future of RPA, the terms “cognitive computing,” “big data,” “machine learning,” and “Industry 4.0” aren’t just buzzwords—they’re critical developments in progress which RPA is gearing up to redefine. The past few years have seen RPA’s meteoric rise as it’s disrupted the insurance, healthcare, and financial-service industries, but just like the evolution of industrial robots, it’s clear that RPA will find more uses and integration across nearly all industries as it continues to develop.
Intelligent Automation has often been described as a continuum starting from process-driven RPA and progressing all the way to data-driven AI. However, the lack of a common, official definition of artificial intelligence has often led to confusion, particularly among consumers attempting to choose the solution that will best meet their needs.
Artificial Intelligence (AI) can be defined as “a branch of computer science dealing with the simulation of intelligent behavior in computers” or simply “the capability of a machine to imitate intelligent human behavior,” according to Merriam Webster. Conversely, Robotic Process Automation (RPA) refers to “software tools that partially or fully automate human activities that are manual, rule-based, and repetitive,” as described by the Association for Intelligent Information Management.
From these two definitions we can find a distinction between simulation of human intelligence (as in the case of AI) and the programming of software to perform manual tasks as denoted in RPA. Within the field of AI are other forms or subsets of intelligent automation that should be noted. Machine learning refers to the ability of a machine or system to learn from experience. In other words, the ability to arrive at a conclusion about data without receiving specific instructions for how to do so. This is achieved through the use of algorithms. Deep learning is quite similar to machine learning but is typically used for much larger data sets.
Choosing between robotic process automation and artificial intelligence comes down to a number of factors such as the parameters, complexity and goals of the project. In attempting to navigate between these two approaches, it is helpful to have concrete examples and real-world applications to compare. This article will provide examples and compare and contrast use cases of RPA and AI within the insurance and financial industries.
With a reported 250 robots in use, BNY Mellon has been implementing robotic process automation since 2016 to streamline its flow of operations. For trade settlements, a number tasks must be completed such as resolving discrepancies, clearing trades and conducting research. Reconciling a failed trade would normally take 5 to 10 minutes, however, RPA was able to complete the task in “a quarter of a second.” The bank claimed that general processing times where “bots” were implemented achieved an 88 percent improvement. Similarly a 100 percent accuracy rate was reported across five systems for account-closure validation processes.
To gain a better understanding of why RPA has proven to be a good fit for BNY Mellon and its performance goals, it is helpful to have a quick overview of the trade settlement process. In a paper published by the Association for Financial Markets in Europe (AFME), we note the following definitions:
From these definitions we recognize some of the routine exchanges between buyer and seller and how RPA could be programmed to handle parts of the trade settlement process, particularly repetitive and manual tasks.
Now, let’s take our trade settlement example a step further. What if a settlement transaction is unsuccessful? How could automation be useful in helping to determine potential causes and also strategies for avoiding failures in the future? These are complex inquiries where a certain amount of analysis would be required. As such, artificial intelligence could be a useful tool.
According to research published by the Board of Governors of the Federal Reserve System, failed transactions or “settlement fails” as they are called in the financial industry, have been found to be systemic within certain time periods. Efforts to curtail instances of settlement fails have been a priority among policymakers and market groups. This is due to the fact that “large and protracted settlement fails are believed to undermine the liquidity and well-functioning of securities markets.”
Due to the often systemic nature of settlement fails, a large amount of data would naturally be generated and recorded from these transactions. These data repositories if properly cleaned (checked for errors) could serve as the training data for an AI or machine learning solution. By analyzing the data set which may include categories such as settlement date, trade date, or currency type, the machine learning algorithms could identify distinctive patterns. As a result, these predictive algorithms could be applied to new data sets to predict the risk of settlement fail before it actually occurs.
There is evidence of increasing implementation of robotics process automation in the insurance industry. Several key areas in the insurance market may be well-suited for automation include claims, product underwriting and pricing and policy administration and servicing.
For example, claims processing often requires a substantial amount of staff time and coordination of documents. In a case study, global RPA software supplier Kofax described how it assisted a client with automating its claims processing routine in an effort to achieve greater efficiency.
“We sell a large proportion of finance and insurance products through car dealerships, which routinely rely on pen and paper. Claims processing was a similarly paper heavy process. In each of those areas we have to collect a significant amount of documentation from external stakeholders and customers. Our administrative teams spent considerable time and effort scanning and processing hardcopy documents, as well as reviewing and sorting email attachments into folders.” -Tim Dewey, former Vice President of Operations Technology, Safe-Guard
To better understand why RPA was a good fit for Safe-Guard’s operations, we will explore the stages of how this process was carried out. The company began by implementing a selection of automated solutions from Kofax’s product suite. An initial step for automation of the claims process was the scanning and uploading of contract and claims documentation to a central data repository.
The bots are also programmed to distinguish the source of documents (such as email vs. fax) which accelerates the process of organizing the large amounts of paperwork. These documents are transported to a holding station or work queue prior to continuing in the workflow process.
The next step entails automated extraction where information of interest (as established by Safe-Guard) are extracted from the various documents and then stored in the system. To track the effectiveness of the automation process, opportunities for improvement and areas of efficiency are monitored.
The software developer reported a significant reduction in the time required to capture details from documents. Specifically, this translated to a reduction from roughly two hours to 10 or 15 minutes. Staff tended to handle documents as much as five times but RPA was able to reduce it to a single handling. Other improvements were noted in areas such as productivity and customer satisfaction increasing by 30 and 15 percent, respectively.
One of the biggest challenges in the Insurance industry is combatting instances of fraud. There are significant financial implications totaling more than $40 billion each year according to the Federal Bureau of Investigation.
How could automation be used to detect or help prevent fraud? Since a combination of pattern recognition and predictive analytics would be needed in this scenario, An AI solution could possibly provide useful insights. For example, companies such as H20.ai which has developed a machine learning platform with a reported 14,000 client organizations, provide AI solutions for the insurance industry.
After undergoing a training process using data from insurance fraud cases, a machine learning algorithm would be primed to recognize patterns linked to fraud risk. Using a probability scale, claims with low or minimal risk would be advance for processing while those with higher risk would be flagged for review. One may also envision an AI assisted RPA solution, where the initial automation can be achieved through basic programming and the analysis of risky claims would be supported by AI.
AI models could provide insights into why claim denials by pinpointing the main aspects that a human analysts should focus on. Targeting these areas helps to minimize labor, accelerate processing times and reduce errors. Customers may also find these solutions helpful as they can be informed of issues with their claim before they are submitted for processing. This helps reduce rates of fraud by streamlining the overall process for a more efficient workflow.
The RPA market is still emerging and evolving, so competition among software developers and providers is intense. This provides you with choice—and decisions to make. When comparing RPA software for your business, pay attention to technology features (such as the ability to work assisted and unassisted), ease-of-use/UI (user interface), security, integrations, and vendor support.
Currently, the top three RPA software and tool providers for business are:
Other major players include OpenSpan, Work Fusion, Kofax, and EdgeVerve.
Let’s see how the major offerings stack up:
Blue Prism offers one of the most complete packages of robotic process automation software. That’s not too surprising, considering that the term “RPA” was coined in their own labs. Replete with enterprise features including security, and a well-thought-out operating model, Blue Prism provides a sophisticated level of customization and depth of implementation for complex user processes. Some of their partnerships include NHS, Accenture, Haxaware, Hewlett Packard Enterprise, Capgemini, and IBM.
UiPath has a highly customizable and extensive architecture, infused with features like predictive analytics, open-platform customization, and a sleek interface. This combination, along with its vendor support (as with Citrix), has earned them an excellent track record for automation support and security. UiPath RPA software supports some of today’s business giants like SAP, Ernst & Young, J.P. Morgan, Deloitte, and the BBC.
Automation Anywhere is a pioneer in the intelligent digital workforce and one of the largest providers of RPA software across the U.S. As one of the most powerful RPA tools with one of the easiest-to-use interfaces, Automation Anywhere boasts partners such as EMC2, KPMG, Genpact, and Infosys.
As we’d hinted earlier, RPA software bots can be broken down into two basic categories: attended and unattended. They’re both similar: Each can bring greater automation and efficiency to a system. But “attended” and “unattended” bots operate according different definable parameters. Importantly, the two types of bots perform different tasks with varying amounts of required human interaction.
Let’s examine this distinction a little more closely:
Attended RPA works alongside the user, helping him or her to incorporate automation into a specific directed task. In this instance, the attended bot is usually constrained to the programs running on a single workstation at a time.
An attended RPA bot requires a human to tell it to start working. For example, let’s say that Sally, one of your employees, needs to scour her in-box for every email with a specific subject, and then copy the details from that email and paste/record it into a separate program. With her attended bot at her side, Sally could simply tell the bot to “process this,” and the bot would dutifully complete the entire grueling task in an instant.
Call-center and customer-support operators can also make efficient use of attended RPA on a wide scale. That’s because these positions often require quickly switching between multiple programs and screens to retrieve information while talking on the phone with a customer. Attended RPA bots can help these employees interface with and retrieve data from any number of applications; they can also accomplish these tasks far faster and more efficiently, freeing the reps to focus more on the customer.
While attended automation is well-suited to tasks that require human-to-system interaction in real time, unattended RPA is far more efficient for all other systems.
Unattended RPA doesn’t require user input or attention, once the program is set up to execute. The only human intervention required would be to evaluate and/or change the directed task, since these bots work 24/7.
Think of Sally from our earlier example. When it comes to scrutinizing incoming emails, finding the right data within them, and then pasting that data into another system, Sally could derive even more benefit from an unattended RPA tool. This bot could effectively manage her email account entirely on its own, watching for and transcribing incoming data from emails with specific subject lines—and then immediately sending that data onward.
Unattended RPA tools allow businesses to optimize a multitude of tasks that have traditionally been error-prone and time- and energy-intensive, due to their reliance on human interaction. Thanks to the adaptable nature of unattended RPA tools, businesses across a wider array of markets are taking advantage of this software to do things like:
Unattended RPA’s higher level of automation makes it an enticing candidate for most businesses. Still, the strict rules and structure of digital information required to guide each bot make this software impractical for many complex applications. Therefore, many businesses are finding that the optimal solution for increasing productivity entails a balanced combination of attended and unattended RPA tools. But if the history of RPA and the rapid development of AI are any guide, it seems all but inevitable that nearly all RPA will become unattended, not long from now.
Additional helpful industry specific resources to inform you along your robotic process automation journey can be found below:
30% to 50% of RPA implementations fail because companies focus too much on the technology itself, instead digging deep enough into processes, at the individual work activity level, to determine the optimal robotics use cases. Just because you can put RPA on a process, doesn’t necessarily mean you should.
Understandably, many businesses are not even aware that RPA exists to make their processes – and therefore the people who form the business – more efficient. Think of it as a no-code automation solution that makes everyone’s working days easier by reducing the monotonous and routine data movement tasks they get hung up on.
In fact, in the 2017 Randstad Employer Brand Research survey, 40% of respondents said they think automation would make their jobs easier and an impressive 72% said they didn’t mind retraining for job changes as long as their pay remained the same or increased.
To ensure that your RPA implementation goes according to plan, follow the steps outlined in the four phases below to show your employees and leadership why RPA helps them, not replaces them.
Determine what processes are core to your business, what processes are secondary, and what processes are redundant.
Generally, the top processes that businesses benefit from automating are:
As you begin to narrow and refine the scope for your project, focus on processes that are repetitive in nature and performed by the largest scale group in your company to maximize the return on investment.
These should be processes should be further narrowed down into those that currently take up most of your employees’ days such as;
One telling metric is to look at the error rates of each department and determine where those errors are happening; often it’s in processes that involve manual data movement that could be automated with RPA instead.
There are a number of ways to identify which routine processes are causing the most trouble or taking up the most time. If a department has been struggling with meeting KPIs, they could benefit from automation, which is able to perform tasks much quicker than human workers and can work around the clock.
By automating these actions, you increase efficiency and spare employees from the dull clerical work they don’t enjoy.
Contact us now to speed up your RPA analysis, use case development, and implementation.