This is the first blog post of a series about RPA – SAP Intelligent RPA and the Next Level of SAP Intelligent RPA.
Back in the 1990s, first screen-scraping technologies and macro technologies came up, that later should be incorporated into a framework we call today Robotic Process Automation – RPA. First RPA tools came up in the 2000s. These predecessors of today’s RPA tools were spot solutions, and the SAP-acquired Contextor was one of those pioneers. They focused on the automation of a limited set of repetitive tasks in their individual industry-specific solution. Only around 2015, RPA became mainstream, and with that also its today’s name. At the same time, also the RPA platforms became broad and generic to target the automation of “any task”.
Today, RPA solutions deliver that promise: they automate from old mainframe applications to classic Windows applications as well as modern web-based applications. Direct API accesses as well as being able to invoke data access protocols like SQL are also more and more common in modern RPA tools. RPA started to merge with other technologies like Machine Learning, we call this Intelligent RPA then.
Main use case areas are still, like in the beginning of RPA, highly repetitive tasks, where the highest benefits of automation exist. In the ERP environment, those are often processes from Shared Service Centers, like in Finance, Procurement or HR.
Looking at the overall RPA customer base, it continues to be a rising technology. After Gartner announced in 2019 that it is in fact the fastest growing segment in the ERP market, the growth in 2020 is also expected to continue, despite the ongoing global challenges.
There are now a few trends which are worthwhile to look at further in this blog post.
RPA is here to stay
First things first: RPA is here to stay. We can call it a universal constant of enterprise IT landscapes: there is always at least one legacy system that no one wants to touch. A migration of this “one system” to a modern replacement would cost thousands or even millions, and somehow this system is still critical in the businesses operations. Making sure we integrate and automate old and new and everchanging heterogenous enterprise IT system landscapes, RPA is the power-tool for the job. The light-weighted approach of RPA, applying efficiency gains on-top of any existing systems, by leaving them untouched, is the undisputed differentiator to any other automation technology.
Furthermore, RPA is implemented quickly: RPA bots can be created in just a 2 weeks project. A good example here is a project for the administration of the swiss canton of Zurich to automate applications for public funds regarding short-term working arrangements, that includes an AS400 system.
RPA to become one service in a business technology platform
Let’s face it: The real goal is enterprise efficiency. Running the daily business as cost-, resource- and time-efficient as possible. Truly, RPA is an important component in this endeavor, and RPA will stay. But: RPA is not the only tool in the toolbox. For reaching end-to-end automation, combining RPA with further technologies and measures allows a next level of efficiency gains.
One of the widely used combination is RPA together with OCR or other document information extraction technologies. As many RPA use cases process unstructured documents (like invoices), and target to enter the information of these components into an ERP system, OCR and similar technologies are important.
RPA bots are usually short runners, whereas business processes that also include manual steps can be truly long running tasks (from a few days to a few weeks). BPM targets modelling of stable and clean workflows - so why not combining them? Let’s take the customer onboarding process, adopted for an online bank. A BPM workflow models the approval steps and orchestrates the full process, RPA bots are steps in this workflow to automate unstructured tasks, like reaching out to a credit check department.
Combining RPA with Machine Learning (ML)/Artificial Intelligence (AI) allows to train complex models to take decisions that cannot simply be modelled with a rule framework. Plugging in an ML Model as “decision engine” into an RPA bot unleashes new potential. Another option is using a chatbot as NLP interface to trigger RPA bots in the background and execute the end users requests.
Finally: RPA is an integration technology, combining RPA with integration middleware technologies, application extension frameworks etc. helps to bring the enterprise IT landscape together.
Why does this matter? There are 2 things to consider:
Even if we would like to have RPA being a “business unit” tool, where central IT involvement is limited, the moment we automate sensitive processes, we need central IT support for security architecture, access governance, auditability and the very least: high availability. Imagine your RPA landscape breaks during your financial closing period and your bots are supposed to do 50% of the work. That leads to the situation that the aim is to have a harmonized IT landscape, where overall maintenance and operations can be done altogether by an IT organization. Thus, RPA and complementary platforms are to be integrated into an enterprise platform.
RPA can only be as efficient, as close and integrated it is with the applications it is supposed to automate. Surely, technologies like surface automation allow us to automate applications that run inside a virtualized desktop environment, but this integration is loose and might cause regular maintenance efforts of the RPA bot. A tight integration between RPA and the enterprise applications allows stability, performance and bots that run for a long time without having to be “fixed” every few weeks. Here, RPA again is required to be “one” with the enterprise applications.
Low-Code/No Code - a global business technology trend also affecting RPA
Low-Code/No-Code (LCNC) development platforms are another emerging trend in the enterprise technology world, that should not be ignored. As the speed of change and the need for agility rises, the issue of finding the right talent for doing complex tasks in companies rises. Simplifying the development or extension process of enterprise applications without having to regularly hire the latest top talent from the markets becomes unavoidable. Allowing this without having to write programming code, can ease the limited expert resource need, as more members of an organization can do the job of bot building. Of course, not the last complex task can be done with LCNC platforms, but certainly a good portion of it, and that is why the trend exists.
RPA cannot shut its eyes in front of Low-Code/No-Code development experience. Also, in RPA we ideally want the members of a business unit to build the bots, as they know best what the steps are, that the bot should automate. Having graphical development tools is an easy start of building the first bot. Also, having the help of a large community and a wide set of online resources is one ongoing trend for RPA.
This of course also affects the complementary technologies to RPA, e.g. Ruum where a LCNC RPA platform will fit well to a lightweight/no code workflow tool.
Intelligent RPA beyond AI services
ML and AI can be used in RPA for more than being components of the bots for taking decisions or processing unstructured data. ML technologies can be an integral part of RPA itself.
A good usage is leveraging computer vision technology to make surface automation more stable. When we cannot integrate into the UI or application protocol, as the underlying application runs inside a virtualized desktop the classical way was to remember the pixel position where the screen elements are and invoke input device activities (like a mouse click). That this breaks regularly is clear. Why not training a CV model to recognize screen elements, as described in this blog post, and then actually detecting safely the button to click, the field to fill, etc.
Building the bots is a simple task, yet it could be improved even further. Already the data-driven insights on what are task hotspots to automate is golden information. Here already today, process mining and RPA go hand in hand. In the future, user behavior mining, just like in Spotlight, can track how UIs are used and gives us the right data basis to have an AI-assisted bot building experience. This might still take some time until “the system that automates itself” becomes somewhat close to reality, but the trend is visible.
Surely there are more trends that will affect the world of business process automation, nevertheless we see those discussed in this blog post as impactful ones:
RPA to be a component that will always have its place and value in enterprise IT landscapes
RPA being limited standalone, but as an integrated component it becomes an extremely powerful and stable business technology
Low-Code/No-Code development experience is a must-have
ML and AI really should be infused into RPA itself, and not only provide services in a bot's workflow.
Please see future developments of Intelligent RPA in the next blog post of this series "Bringing SAP Intelligent RPA to the Next Level" by andreas.gerber. Also stay tuned for more exciting news around RPA here on SAP Community in the following weeks.