Robotic Process Automation: a modest step into the future

As many of you will recognize, the current crisis has prevented many of my usual leisure activities from happening. However, this has allowed me a measure of spare time which I have used productively. Last month I achieved my certificate from UIPath, the leading tool in Robotic Process Automation (RPA). I may now call myself an advanced RPA developer. In this short article I would like to explain what RPA is and what its applications are. I would also like to discuss some possible pitfalls.

Making sense of the buzz

Over the last few years there has been a lot of talk about robots and Artificial Intelligence (AI). What used to be science fiction has become a part of everyday life. As I wrote in my previous article, voice assistants have become common in households and are set to arrive in the business world as well. Another great innovation of the past few years is machine learning. In machine learning, a piece of software does not only follow instructions, but also experiments in order to improve its performance. Due to machine learning, great progress has been made in fields such as Optical Character Recognition (OCR) in which a computer can optically recognize text from a picture. RPA has joined this toolset as a way to automate office tasks more easily.

RPA means using a robot to automate a business process. Examples include processing invoices, doing the paperwork when a new employee joins the company or sending out business reports. These robots will not engage in machine learning. So, it is not possible to tell a robot to “handle incoming invoices” and then sit back and relax. The robots that are engaged in processes can be compared to the robots that are able to paint a car. The robot needs to be specifically built and instructed to do a certain task. However, other elements of AI have been key in making RPA possible in the first place. OCR is an integral part of RPA. This is because the great advantage of RPA is that it is independent of the kind of tool you are using. Many business tools already provide the user with the possibility of automating its processes. For instance, some years ago I built a tool for the TNT-FedEx integration in Excel, using visual basic. While back then (this was in 2016) I was able to automate the entire Excel process, we still needed two administrators to handle the interfacing with human resources and enterprise resource planning systems. RPA allows you to automate an entire process, regardless of the tools involved. This means that an entire process can be automated with no manual labor needed to interface with other tools.

So while robotics has had a tremendous impact on industrial production already, it is set to have an increasing impact on the office world as well. Building an interface between tools is a notoriously difficult and expensive endeavor. RPA can help to bridge the gap between legacy software and a more automated new solution. In addition to this it allows a user to run products from different vendors easily, using RPA as the interface. Hereby systems that would otherwise be incompatible are linked up.

Obstacles to a robotic future

To me the idea of RPA sounds great! The benefits are clear and the potential savings tremendous. Organizations can automate the more mundane tasks to free up their often highly educated workforce to focus on more strategic and creative activities. RPA is growing: the total revenue of RPA providers grew by 63% in 2018 and another 50% (expected) in 2019. Thought-leaders such as Gartner expect significant growth in 2020. In their latest funding round in April 2019 UIPath was valued at $7 billion. Competitors include Automation Anywhere which focusses on equipping its robots with machine learning capabilities and Blue Prism which has been pioneering in the RPA field since 2001. However, the total revenue within RPA is still small, estimated at $1,3 billion. After working with the RPA tool UIPath for a while, I personally see three obstacles:

1.     It is too hard to use for the average user

UIPath is trying to sell their solution as something which business users will be able to use themselves after a training. I believe they are overestimating the programming skills of most business professionals who work in human resources, finance or other non-IT fields. On the other hand, the development can also not be handled by programmers alone. Successfully automating a process will require either someone who is a subject matter expert or a consultant both skilled in business processes and  programming.

2.     It requires intensive maintenance

As someone who works in the private business sector I know that there is no such thing as stability. Changes to the process will be required on a regular basis, at which point the same skillset is going to be needed as for the initial implementation. Although a much lower time investment will be required for tweaking a process already in place, the business will need to be able to call on this expertise at any time.

3.     Licensing is unclear and too costly

After I achieved my certification I started to look into licensing costs to see how easy it would be to implement RPA solutions. I was shocked by the absence of clear pricing. UIPath and other vendors supply a community edition for developers to experiment with and automate some processes. Any use on an enterprise scale leads to high startup costs as well as dependency on the vendor. This makes RPA less attractive, especially to smaller companies.

The bottom line

In sum I would say that RPA has enormous potential for the future once these hurdles are cleared. In addition to the leading vendors there are some promising open source software projects. I believe that  the current leaders will be outcompeted by a combination of open source software and consultants who will implement RPA and then maintain the processes at a reasonable cost. This way the Return on Investment of a RPA implementation would be calculated more easily. Simply investigate the process to see how many working hours can be saved, and compare this with the costs of implementation and maintenance. Although this in itself will not lead us into the future it may free up some of your colleagues’ time, so they can focus on designing nuclear fusion engines, underwater cities and hoverboards.