20 November 2019
Digital Robotics — also known as Robotic Process Automation (RPA) — is a new technology with the potential to radically increase business process productivity by eliminating repetitive manual work and media breaks. At the same time it improves data quality by executing processes more consistently than humans.
According to RPA software vendors, digital robotics is the new silver bullet for process automation that generates sustainable business value and provides rapid return on investment. But does it really deliver on these promises?
It is not black and white
Let us take the top 5 claims and examine them a bit closer. In our analysis, we studied online documentation and reviews, tested multiple RPA tools with simple workflows and implemented a few real-life business processes using a leading tool. And we constantly compared what we learnt with our multi-decade experience in the IT industry.
1. Digital robots are fast to build and deploy
True: with a little practice, one can pull together automated processes quite quickly. Within a few days, even business processes of middle size complexity — such as an accounts receivables creation process — can be implemented. This is remarkable, especially if we compare this with the typical duration and cost of traditional IT automation projects where business applications with various interfaces need to be built and integrated.
2. Digital robots are cheap
Partially true: as stated above, creating a digital robot is done relatively quickly and therefore is cheap. However, version 1.0 is just the beginning of the robot's lifecycle. With time, its execution environment will change — and the robot needs to be adapted, too. Even small changes in the underlying software infrastructure (e.g., a Microsoft Office update) or a modified input file can render a digital robot useless. Thus, continuous maintenance and adoption of robots is an inevitable activity and can become costly over time.
Of course, the robot could be built to be resilient to changes in its execution environment — just like enterprise software should be fault-tolerant. However, it would considerably increase the cost to initially build the robot since it has to be able to cope with a wide variety of inputs. Sustainability and future-readiness take their toll.
3. Building digital robots does not require programming skills
Partially true: RPA vendors target business departments with the claim that digital robots can be built without programming skills. To fulfill the claim, they offer easy-to-use tools such as drag and drop designers and process recorders. It is true that the modern drag and drop graphical user interfaces (GUIs) simplify robot design but they do not remove the necessity to understand basic programming concepts. For example, dealing with sequences, loops, parameters and variables cannot be avoided.
"But the product's recording tools can really be used without programming skills, right?" - you might ask.
Unfortunately, recording tools are far from being perfect. Determining the real intention behind a mouse movement or a click is challenging: some recorders misinterpret the flow of user actions and the generated script needs to be manually corrected afterwards. Others generate automation scripts that fail already if the window position has slightly changed compared to the recording, making a complete manual rewrite necessary. In both cases, basic programming skills are required to succeed.
4. Anything can be automated with digital robots
Not really: it is a recurring pattern in software vendors' communication to position their product as the solution for all automation problems. Reality is more complex, though. Digital robots can be quickly deployed to automate simple processes or bridge media breaks — at the same time they are generally sensitive to changes in the environment.
Complex processes often pose challenges to RPA solutions that these cannot master. Typically, processes with a higher variety of inputs and alternative flows, with asynchronous processing and potential unexpected events or with a complex IT environment can quickly overstrain a digital robot. Last but not least, we have seen issues with web applications where robots have difficulties to identify dynamically generated HTML page elements repetitively and reliably. Digital robotics definitely has its boundaries.
5. Digital robotics comes with intelligence to automate human decisions
Not really: RPA is used for automating repetitive manual work, ultimately freeing up valuable capacity of business experts. When running, digital robots follow strict instructions and rules — this cannot be called intelligent. Any unforeseen event (e.g., inputs not fitting to the scheme, systems not responding etc.) will cause the digital robot to stop and ask for human help. Changes in the environment might also prevent the robots from completing their work. In these cases human intervention is required.
To decrease the number of such events, some RPA solutions utilize advanced tools to interpret ambiguous inputs such as machine learning assisted image recognition or natural language processing. The effectivity of such tools depends heavily on the use case and the reference data the underlying algorithms are trained with (aka training data). Even in the best case, unexpected events cannot be eliminated completely. Let us take the simple example of an "intelligent" robot classifying customer emails: in this case, the robot examines each incoming email and tries to assign them to a category — based on its training set of historic emails and classification decisions. It will work fairly well as long as the incoming mails show great similarities to the training set. Anything new or different might land in a wrong category or left unclassified.
Machine learning does not make robots intelligent: it is just a more sophisticated and complex way to interpret input data and to make process decisions. Digital robots still need human monitoring and assistance in their daily operation.
How Digital Robotics can make sense
Digital robotics provides some interesting benefits but is not the universal solution to automation requirements. It has to be well thought through what to automate with RPA and what to tackle with other approaches (such as traditional application development). Processes with a low frequency of related IT changes are ideal candidates for automation with digital robotics since long-term maintenance costs do not overshadow the efficiency gains. So are one-off data migration initiatives, as the short-time added value is very high while the expected robot lifetime is short. Automating processes in ever changing IT-environments require more robust and failure-tolerant tools to achieve the business value hoped for.
Digital robotics is no silver bullet: it is a valuable tool in a greater toolbox. It should be evaluated for each automation requirement, which tool is the most effective and efficient to implement it. In some cases, RPA will turn out to be the best answer — in others, it will not provide a sustainable and economic solution. How can we be prepared for upcoming automation requests? A good approach is to define a technology inventory (i.e. the toolbox) and a decision guide that can be applied to simplify and speed up the technology selection process. Additionally, if you want to add digital robotics to your toolbox, try it out and run a few proof of concept projects. Have fun!