It’s easy to get robotic process automation (RPA), machine learning (ML), and artificial intelligence (AI) mixed up. People use them interchangeably and it can be hard to differentiate between the three when they’re flying around in conversation. But they’re not as mystical or opaque as they seem. You benefit from them every day when you ask Alexa to set a timer, watch a movie from your Netflix recommendations, or break down and order those footie pajamas that Amazon has been recommending to you for the last two weeks (just me, or…?).
Hopefully, this little crash course will give you a better understanding of the differences between these technologies and how they apply to process automation.
Let’s get started!
As I explained in my post on RPA, robotic process automation is software that helps automate a manual process. Unlike ML and AI, which are data-driven, RPA is process-driven. By process-driven, I mean that RPA works via pre-built processes. These processes are repetitive, rule-based, and usually require a human to interact with more than one line-of-business system (such as SharePoint or Office 365). RPA then automates these processes for the human end-user. Think of it as RPA being the brawn and ML and AI being the brains.
When it comes to process automation, RPA does a great job with tasks like invoice processing. Normally, an employee downloads electronic invoices from their email and creates bills in their accounting software. An RPA robot can help automate the manual parts of this process (retrieving the invoices, downloading them and creating them). However, it would require some help from machine learning to finish the job. Machine learning would need to intelligently “read” the invoices and extract the required information from them (supplier name, invoice number, due date, and any other information it has been asked to extract) before handing the job back to RPA to create the invoices in the system.
Why is this? Well, RPA isn’t capable of intelligent “thought” like ML and AI are. A human must explicitly program every process so that the bot doing the automating knows exactly how to carry out the task. This means RPA is best when automating rule-based, highly repetitive tasks.
Click to Tweet: RPA isn’t capable of intelligent “thought” like ML and AI is; that means RPA is best used for automating rule-based, highly repetitive tasks.
Machine learning (ML) is an application of artificial intelligence that enables systems to learn from data without being explicitly programmed. ML is based on the premise that we can build technology that can process data and learn from the data on its own, without the constant supervision of programmers. It aims to learn from data, improving accuracy as it learns. Spotify uses machine learning to create your Discover Weekly playlist every Monday. As you stream music, Spotify’s machine learning algorithms use your data (i.e. the songs you listen to) to create a weekly two-hour-long playlist of music it thinks you’ll enjoy.
Netflix, YouTube, Amazon, and other services use ML to do this, too. They rely on viewing or browsing history to recommend content or products they think you’ll like. Have you ever noticed that services like these seem to know you as well as—or even better than—you know yourself? That’s because the more you use services like Spotify or Amazon, the more they learn about you. The more they learn, the more accurate their recommendations become. This is machine learning in action.
In process automation, machine learning can find like documents, identify what they are by using image recognition, and sort them under the correct classification. As you can imagine, classification is much faster with machine learning than when done manually by an employee.
Finally, we’ve reached the most misunderstood of the three terms.
Although artificial intelligence can seem like magic, behind the scenes, there’s really nothing mystical about it. At a high level, AI is essentially an umbrella term for various software—such as machine learning, as I mentioned earlier—that can demonstrate intelligence. Unlike RPA, AI is driven by data. And while machine learning aims to acquire knowledge, AI actually aims to become more intelligent. Its goal is to simulate intelligence. Most recently, AI’s been in the news for its use in creating FaceApp, the app that uses AI technology to produce a creepily realistic aged version of a photo of your face. But probably the most well-known example of AI today is Sophia, the human-like robot created by Hanson Robotics in 2016. Thanks to AI, Sophia can (for the most part) hold conversations and (for the most part) act like a real person.
Hopefully, this helped you better differentiate between RPA, ML and AI. All three have very useful business process applications and can be enormously helpful in improving productivity. Now go out and flaunt your newfound knowledge. And don’t think of Sophia trying to make human facial expressions.