“I have not failed. I've just found 10,000 ways that won't work.”
—Thomas A. Edison
Since robotic process automation (RPA) hit the marketplace hard in 2015, there’s been a lot of complaining that it doesn’t seem to work. This perception is neither accurate nor fair. But given the enormous promise of intelligent automation, and its inevitability, why do so many RPA projects fail?
Most of the RPA project failures I've seen were highly subjective, and not very extreme (You'll find the details in my new book, The Care and Feeding of Bots: An Owner’s Manual for Robotic Process Automation). Many so-called RPA failures are more like graduating from high school with a 3.0 grade point average, as opposed to not graduating at all.
This is not to say RPA hasn’t seen its share of spectacular failures. But, by most measures in the information technology world, RPA fails have been fast and small—and that's part of RPA's attraction. If your IT project is going to fail, isn’t it better when it’s a $100,000 RPA pilot, rather than a $100 million ERP upgrade?
So how you do succeed with RPA and AI? You need to understand all of the different ways I’ve seen these technologies fail over the last decade or so. By understanding these failures, you'll know what you need to do to avoid repeating them in your own efforts.
The 5 types of RPA project fails
In mid-2019, if you googled the phrase “RPA implementation failure,” you’d find over 5 million hits. This is not a ringing endorsement of this technology. But context is important, and in this case, it is necessary to unpack the word “failure.” What exactly is meant by “failure” when describing a typical RPA implementation?
To make sense of why RPA projects fail, you need to classify the types of failures that occur. These categories, in descending order of frequency and intensity, include:
- Financial—Failure to deliver business value or meet expectations
- Governance—Failures due to errors in management and oversight
- Operational—Failures due to operational errors or missed operational expectations
- Design—Failures due to poor design practices or principles
- Technical—Failures of the software to perform as expected or needed
Let’s review each of these in turn to understand how and why RPA seems to have such poor performance in practice.
Financial failure
The vast majority of RPA failures are financial in that they fail to deliver the expected value. This is nearly always due to a lack of expected cost savings, since few RPA implementations focus on revenue generation. In these implementations, the bots themselves might have worked perfectly, and often do, but the results did not meet expectations of value and financial returns.
I estimate that this type of failure accounts for approximately 60% of failed implementations—by far the most frequent, and frustrating. Over and over, I have spoken with business leaders whose RPA initiatives stalled after one or two implementations because "the value just wasn’t there.” For a variety of reasons, the benefits provided by RPA do not meet initial expectations, and they do not do so most of the time.
Given that the attraction of RPA is that it can reduce costs, financial failures are particularly annoying both to the companies using the technology and the companies selling it.
Governance failure
Governance failures, those caused by not using bots correctly or effectively, are the second most common cause of RPA failures, representing around 15% of all failed projects. Bots are a workforce that, just like any other, must be managed and governed to maximize their effectiveness.
Most organizations devote significant time, money, and energy to building out a center of excellence (CoE) for RPA. In fact, using a CoE is practically religious dogma in the industry. But if there’s so much attention focused on governance, why would it be the second most prevalent cause of failures?
There are two primary reasons for this.
- Focusing too much energy and resources on governance too early in the process makes RPA overly complex, expensive, and difficult.
- Not focusing enough on governance early in the process simply replicates, and then explodes, the out-of-control, free-range-automation problem that RPA was originally intended to address.
The key here is balance; have enough governance to absorb complexity as your bot population expands, but not so much that the cost and complexity make RPA lose its value.
Operational failure
Operational failures occur when the bots don't perform as expected when placed in production. This is distinct from governance failures, which occur when bots aren’t correctly or effectively coordinated. Operational failures deal with how individual bots operate, rather than how the bots coordinate with others.
I see operational failures perhaps 10% of the time. They include such things as:
- Scheduling bots to run at times when the systems they must access are unavailable
- Building bots that require a level of security access that company policy precludes
- Requiring access to systems that cannot be accessed through firewalls or other technical constraints
Operational failures result not from poor design or poor functionality, but from poor planning and governance. And while financial and governance failures aren’t overly costly, operational failures can cause millions of dollars in losses.
Design failure
Design failures occur when the bot was programmed erroneously. The issue could be related to missed requirements, misinterpreted requirements, poorly executed designs, or errors in solution architecture. My unscientific assessment is that design accounts for perhaps 10% of failures.
Despite growing experience with RPA, poor design practices are still prevalent in many organizations. There was such a frantic demand for RPA skills after 2017 that many people were hired without the necessary skills, and many more were put directly into delivery projects with only rudimentary training.
Companies that hired their own staff often didn’t know what qualifications were necessary and were unable to adequately verify the skills of those they hired. Consultancies and staffing firms would claim to have a well-trained and experienced workforce, but as someone who has worked in that world for most of my career, I can confirm that these companies are extremely talented at overstating their own resources and abilities.
Like operational failures, design failures can be very expensive—particularly if you have discounted the importance of testing.
Technical failure
Technical failures occur when the RPA software does not perform as expected or intended. Examples of this include incompatibilities with certain tools, settings, or standards that prevent the underlying tool from fulfilling its intended purpose. This accounts for the remaining 5% of failures.
I have seen technical failures where there was some unexpected incompatibility between tools or systems, or where some feature didn’t work as advertised. But by and large, RPA has few technical failures, because the underlying technology is both basic and mature. Macros are hardly new, screen-scraping has been around for 30 years, and optical character recognition (OCR) has also been around for decades.
The truly new part of RPA is its coordination and governance capabilities, and some of the development tools that allow developers to create bots rapidly. But these functions are enhancements to the underlying technologies, and hence rarely cause functional failures.
Where they failed, you won't
Not all RPA project failures are the same, and most, if not all, are avoidable. RPA was a new technology five years ago. Today, it is rapidly maturing, and the body of experience that implementers and owners have has expanded.
Yes, there is a range of reasons why RPA projects have failed, but all of those are understandable and manageable. Moreover, the urgency to adopt this technology is increasing, not decreasing.
While the list of organizations that have succeeded with RPA projects at scale is small, it is growing. Organizations that achieve early success with this wave of automation will create a structural advantage that their competitors will be hard-pressed to counter.
The 5% of companies that are succeeding with RPA today demonstrate that the technology does work, and these businesses are creating yet another technology arms race that all organizations must join in order to survive, let alone thrive. Yes, achieving full ROI from your RPA investment in one quarter may protect your bonus this year, but successfully deploying bots at scale over the next year or two will protect your job, and perhaps your career.
This may sound like the same hyperbolic rhetoric that the RPA industry has been guilty of all along. It's not. Rather, it is the inevitable result of the technology demonstrating that it can work, when properly applied, and does create game-changing results. In the end, as with the PC, the Internet, and social media, not adopting RPA is not an option. You just have to go about it in the right way.
This story was abstracted from Chris Surdak's new book, The Care and Feeding of Bots: An Owners-Manual For Robotic Process Automation, which covers the reasons behind RPA failures—and best practices for avoiding them—in more depth.
Keep learning
Choose the right ESM tool for your needs. Get up to speed with the our Buyer's Guide to Enterprise Service Management Tools
What will the next generation of enterprise service management tools look like? TechBeacon's Guide to Optimizing Enterprise Service Management offers the insights.
Discover more about IT Operations Monitoring with TechBeacon's Guide.
What's the best way to get your robotic process automation project off the ground? Find out how to choose the right tools—and the right project.
Ready to advance up the IT career ladder? TechBeacon's Careers Topic Center provides expert advice you need to prepare for your next move.
