RPA: Your Robot Colleague

Robotic Process Automation (RPA) is a technology that uses “bots” to automate repetitive tasks using a set of pre-defined rules & workflows, which attempt to mimic human behavior. You can see them as highly specialized colleagues who really enjoy performing one specific task over and over again. When this colleague joins your team, you will have more time on your hands for more complex and strategic problems. However, these bots are not the smartest. They need clear guidance to ensure that the processes are completed correctly.

Mimicing human behavior

What sets RPA’s apart from other scripts is that they are specifically designed to automate human tasks. This includes front-end tasks like interacting with a user interface. This causes them to not disrupt the underlying systems already in place, allowing for seamless cooperation between bot and human across systems. Humans are also prone to making errors. With a well-constructed RPA, accuracy can increase greatly.

As RPAs can mimic tasks targeted at humans, they are very useful for integrating systems which do not have an application programming interface (API). They can comprehend what is on a screen and complete keystrokes. Entering the same information in multiple systems can thus be outsourced to the bot, increasing the cooperation between different departments and speeding up overarching processes.

Types of RPA’s

RPA’s can be divided into three categories: attended, unattended and a hybrid of the two. The main difference lies in how much humans interact with the bot.

Attended RPAs work alongside humans, executing sub-processes of a broader, more dynamic process. They are waiting to be triggered by a human when needed and their workflow can be applied in various scenarios.

Unattended RPAs on the other hand do not require human intervention. They run on a predefined schedule or are triggered when a certain event takes place. These bots are specialized in executing end-to-end processes. Attended RPAs are more like a personal assistant, whilst unattended RPA’s function as a more specialized stand-alone automation engine.

Hybrid solutions can leverage the speed and accuracy of the bot for repetitive tasks, whilst a human overseer identifies complex deviations. In this manner, the human can guide the automation in the right direction based on the information the bot provides.

RPAs can run locally on a desktop, but also in the cloud. Local (on-premises) RPAs are a solid choice when handling sensitive or high-volume data, or when dealing with legacy systems which require front-end interaction. Cloud RPAs on the other hand are great for processes with varying scalability. Cloud RPAs can also more easily capture triggers based on activity in the network and can be deployed more rapidly.

You can image processing invoices, generating reports, updating customer information or ETL (Extract, Transform, Load) processes as good RPA applications. The less deviations are expected in a process, the larger the portion of the workload which can be assigned to the bot.

Good practices

At the end of the day, RPAs are not rational problem-solvers. You give them a set of instructions, and they will follow them without adapting to deviations. It is thus crucial that enough checks are in place to assure that the process is reliable. For example, if the task at hand is filling out a form on a weekly basis, what should happen when a new field is added to the form? Even if the RPA cannot figure out what the new field should contain, measures can be put in place to detect these deviations and halt the process before making mistakes. Letting the RPA log certain milestones of the process helps the user to locate the errors. The user can then be informed of the deviation and adapt the workflow.

It is better to halt the process early than allow the RPA to make mistakes, especially if mistakes can lead to large consequences. They are designed to save time, so you do not want to invest more time in fixing the RPAs errors than if you were to do the entire process by hand.

Another good practice is to divide the task into more general, independent sub-processes. This allows you to easily recycle certain parts of the RPA for similar tasks in other applications. Sub-processes also facilitate clear documentation, easing in the handover of the RPA to less technical colleagues.