INTELLIGENT PROCESS AUTOMATION: THE EVOLUTION OF RPA Davide CentonzeIPA enables an organisation to optimise the productivity of its people, improve efficiency, and reduce the risks associated with business processes 1 Beyond pure robotics Intelligent process automation (IPA) is basically robotic process automation (RPA) powered by “smart technologies”. IPA therefore makes it possible to progress from solutions focused on standard and repetitive tasks to new models based on a machine learning approach, allowing data robots to develop new skills, make judgements and provide feedback. This is because a standard, repetitive, exception-free business model does not exist. Certain operations that are instantaneous in nature for the operator concerned can be carried out efficiently by a software application only with the support of machine learning technologies. In this scenario, intelligent process automation covers a logical and technological continuum based on multiple components of increasing impact and complexity. Robotic Process Automation (RPA) It thereby replaces the operator's role in repetitive and rule-based tasks, reducing execution times and costs. In such cases, the technology executes and monitors only the simple activities for which it has been programmed. Intelligent Process Automation (IPA) Machine learning tools (ML tools): algorithms and machine learning solutions pre-trained to solve a specific decision-making problem (e.g. recognising an image), which can be further specialised to take into account the data in the company's specific context. Machine Learning Platforms (ML Platforms): highly customised solutions designed to address specific decision-making scenarios, requiring the support of specialists who implement and train these from scratch, relying on data and systems specific to the organisation. It is clear that intelligent process automation represents a genuine paradigm shift in terms of management, organisation and implementation in which, for machine learning components, it becomes possible to go beyond the mere concept of programming to actually “training” the machine. The machine does not return output based on clearly defined rules and logic, but instead follows complex decision-making process, which are self-generated on the basis of historical data or data collected in real time. 2 “It takes the robot out of the human” Humans remain the main actors in the organisation and become increasingly critical to the operation of the IPA solutions, being no longer burdened with repetitive tasks and able to leverage their time for the beenefit of value-added activities. These include interpreting the results produced by the automated process, intervening to handle exceptions, and training the “machine” to make decisions that are consistent with developments in the surrounding environment. This reflects the "human in the loop" paradigm, in other words a situation based on cooperation between human and machine, where the former benefits from machine's data processing capabilities while the latter learns incrementally from the human's insights. A risk of “algorithm bias” therefore exists, in that there is a transfer of the human decision-making boundaries to the machines. Overcoming this requires a significant investment to develop the necessary analytical culture, a prerequisite for increasing the dissemination and effectiveness of the solutions involved. Data Driven Machine Learning RobotsIntelligent Process AutomationData Robotics, defined as the set of technologies, techniques and applications required to design and implement a new automation process based on self-learning and artificial intelligence technologies, facilitates the introduction and integration of automation into organisational processes. Thanks to the application of “smart” technologies Intelligent Process Automation guarantees an improvement of Robotic Process Automation, facilitating the evolution from solutions that handle straightforward and recurring tasks, to new paradigms based on machine learning techniques.