In certain cases, RPA, i.e. a software package simulating the interaction with a "real" employee through a graphic interface with all kinds of applications and systems, can drastically increase productivity and reduce the number of errors.
Such robots are able to work simultaneously with several programs, transfer data among them, fill out forms on their own, perform calculations, verify the correctness of "manual" input, and much moreе. Software robots can be relied on to enter primary data into information systems, fill out forms, accept applications, enter counterparties in databases and even chat with clients. Such "digital employee" can take on the most routine, repetitive tasks and perform them almost faultlessly.
Increased speed of business processes, quick return on investment, ease of implementation and scaling down, unbeatable flexibility and compatibility with any applications, all of this is gradually turning RPA into a real driving force behind "digital enterprises." So, this is no wonder that Gartner has estimated that the global RPA market should reach $2.9 billion in 2021.
However, our experience shows that up to 15 such robots might need to be deployed to automate a single business process, whereas the activity of a large company may entail dozens of business processes, and any of them may require at any stage some deviations from the "ideal" model (non-standard application, new software version, non-standard document, etc.) In such case, an operator needs to be involved to make a decision, and upon high workflow, this gives rise to routine tasks necessary to supervise and monitor the robots' actions.
On overly dynamic environment in which RPA must work creates problems. Any change even insignificant for a person (both in the business application interface and the format of incoming data) requires quick reaction and robot reconfiguration.
For an effective implementation of RPA, it is important that business processes be standardized, well documented and well adjusted. Although many advanced management systems are offered in the modern market, the concept of BPMN itself assumes that we deal with "ideal" processes. Unfortunately, this condition is not always fulfilled in real situations.
So, the business community has formulated a clear request for a gradual transition from "local" automation of individual actions even with the use of robot systems or machine learning (ML) and artificial intelligence (AI) to more global systems of intellectual analysis and expertise.
Such system created with the extensive use of ML and AI could not only perform routine operations but also make decisions about what to do in case of deviations from "ideal" models. For example, when the number of customer complaints increase, an intelligent system could conduct an analysis on its own, determine the cause and propose changes that need to be made to the business process.
It should be borne in mind that it takes a lot of time to describe business processes by standard means, and the employees who are assigned this task need to be highly qualified and have considerable experience. We also see here a huge scope for introducing technologies based on MF and AI as they could be used as repositories of ready-made descriptions and standard business process models. This could help small companies solve this problem which is often beyond their capabilities and experience.
Overall, thanks to the development of RPA, ML and AI systems, we can create sufficiently effective "digital employees" that can fulfill routine tasks and increase companies' productivity. But, unfortunately, for now they are only isolated systems operating locally and requiring constant control. They are only for now entry-level employees though "digital" and with the rudiments of intelligence.