Enterprise IT landscapes are complex and not well interconnected. According to a recent Deloitte survey of IT executives commissioned by MuleSoft, an average enterprise runs over 800 applications, of which less than 30% are integrated, leaving the employees to manually interface with the remaining isolated 560 applications.
Especially legacy applications often do not offer an Application Programming Interface (API) access, but even when an API is available, integration often proves less than straightforward. Insufficient documentation is a major contributor. Furthermore, the administrator of the application has to explicitly grant access to the robots, which can become a major impediment if it is managed externally as is the case for many SaaS applications. Finally, creating custom integrations falls outside the low-code domain accessible without specialized software development expertise.
Robotic Process Automation (RPA) has risen to prominence for its promise of connecting applications without software integration, by simply working through the user interface. Despite its shortcomings, this approach creates predictability of outcome, which has enabled an entire ecosystem of RPA vendors and implementation partners to grow at a rate unprecedented for any other process automation technology.
The major drawback of user interface (UI)-based automation is the high level of maintenance required for running a robotic workforce. UIs were not originally designed to be used by machines. Each small change in the layout of elements, even barely noticeable for a human user, has the potential to bring a robot to a screeching halt.
Thus, a rival approach to low-code process automation has been gaining popularity in recent years. Integration Platforms as a Service (iPaaS), representing a natural evolution of the Enterprise Service Bus (ESB) technology also known as Middleware, relies on pre-build API integrations, or connectors, to enable simple workflow automations.
While RPA was born an enterprise technology focused on desktop automation, modern iPaaS has its roots in consumer webservice automation. IFTTT, one of the best-known pioneers in the field, originally offered the automation of Twitter, Facebook and Gmail. Zapier was founded a year later, in 2011, to provide a similar experience to enterprise users. They were soon joined by then already established ESB providers such as TIBCO, Informatica and Boomi, and by Microsoft Flow.
One major practical limitation of iPaaS is the available universe of connectors. While leading providers offer hundreds of integrations, the selection is similar across all vendors and mostly focused on North American web services with a focus on corporate environments. Using iPaaS as a comprehensive low-code automation platform is thus challenging. Even companies which do so at scale, such as Finn Auto in Germany, rely on complementary RPA-based approaches.
The dilemma between reliability and flexibility is well recognized by both RPA and iPaaS leaders.
In March 2021 UiPath acquired Cloud Elements, an API integration platform. Its October feature release mentioned “introducing enterprise-grade API-based automation capabilities”, somewhat ironically given the initial commitment to UI automation reflected in the company’s name.
Similar logic drove TIBCO’s intent to acquire Blue Prism (a bidding contest is ongoing at the time of this publication). Another iPaaS provider, Workato, which recently announced $200m funding at $5.7bn valuation, puts emphasis on its in-built RPA functionality. We will probably see an acquisition in this space by Automation Anywhere as well after its planned IPO.
In our work at F-ONE, we prefer to use API integrations whenever possible to increase the stability of our robots. However, in most situations it is just not practical or feasible, given how much time it takes to get proper documentation and arrange API access. In those cases RPA provides an expedient solution, with benefits of automation outweighing the maintenance costs of the bots. In order to enable distributed automation initiatives, the predictability of outcome independent of the technological landscape is often a decisive factor.
Generally, we observe a convergence between API-based and UI-based automation, which we expect to continue in the coming months. Enterprise automation leaders are well advised to consider the benefits of both approaches in each individual case and make this choice consciously as part of the overarching automation strategy definition.
About the author: Artem Fadin is the founder and managing director of F-ONE Group, an Intelligent Automation solution company. His prior experience includes strategic growth and M&A consulting work for Fortune 500 and middle-market industrial clients as Vice President at Delphi Advisors in Frankfurt.