Organizational buy-in is a key part of successful digital transformation. Here's how to tap the talent and perspective of some key roles as you build your strategy.
Robotic process automation (RPA): How it works
How does robotic process automation software handle repetitive tasks? Does RPA require coding? Where does it fit? Let’s break it down in plain terms
RPA works somewhat like a macro
Bultman compares the basics of how RPA software works to another form of automation: the macro. You are essentially using RPA software to “train” a bot to complete (and reuse) keystrokes and mouse movements in an application UI, the same way that a person would input to complete the same task. (Nintex’s RPA tool, like some of its competitors, is completely drag-and-drop – the user doesn’t need to write any code.)
"A human ‘trains’ the RPA bot in a fashion similar to recording to a macro, recording each step and action in the process being automated,” Bultman explains. “The bot will precisely mimic every mouse movement and typed interaction with all of the systems and applications involved – logging in, copying data from one location and pasting it in another, copying files, etc. – until the process is complete. Once trained, the bot will execute those steps repeatedly, precisely, and at high speed.”
In effect, you’re using RPA software, often via a drag-and-drop tool, to teach a bot to interact with a computer UI in the same manner that you do, using the same credentials. And you’re teaching it to do that over and over again. This is why dynamic or adaptive processes are usually poor fits for RPA. If it’s so mundane as to tread on boring, it might be a good RPA candidate.
“Common tasks, alone or in sequence, for RPA bots are things like opening email and attachments logging into enterprise applications, copying and pasting data, moving files and folders, filling in forms, reading/writing to databases, scraping data from the web, making calculations, collecting social media statistics, extracting structured data from documents, and following if/then paths,” Bultman says.
How to build RPA bots
Huff from Kofax breaks down three fundamental steps to building RPA bots:
1. Identify the right candidate [for] automation and conduct a keystroke mapping session that documents each process, step-by-keystroke, along with decision trees, business rules, and exception-handling rules.
2. Identify where automation will play and where humans will play within the automation, and account for how the automated solution will be orchestrated between RPA and human. “In my experience,” Huff says, “ the most successful and sustainable automations are built with the human in mind and where the value is in empowerment and not arbitrage.”
3. Test and deploy the automation while monitoring for required calibrations.
Step three might be easily overlooked, especially in the early phases of RPA projects. “Automatic” has its limits, in a sense – RPA isn’t actually a set-it-and-forget-it tool, and you’ll need to monitor and measure the right things to ensure success. This is an area where the ease-of-use that many RPA software tools strive for could backfire; it may make it seem like RPA itself is easy. But Huff and other experts stress the importance of continuous optimization.
“RPA at scale should receive the same continuous process improvement attention that manual processes receive from Six Sigma analysts,” Huff says.
“The people, processes, and tools to actively govern RPA should be deployed through a sustainment model that’s often referred to as a Digital Management Office or Center of Excellence (COE). A holistic COE should have at least six competencies and include Strategy & Governance, Tools & Training, Performance Measurement & Reporting, Innovation, Change Management, and Operations.”
[ How can automation free up more staff time for innovation? Get the free eBook: Managing IT with Automation. ]