RPA vs Machine Learning: The Origin Story of AI

RPA vs. machine learning: The origin story of AI

Robotic process automation (RPA) vs Machine Learning (ML): they are both subsets of artificial intelligence (AI), but what are they really? How do they differ? And how are they being used today?

To understand the current power and future potential of these technologies, it can be helpful to go back to the beginning. By knowing how they developed, we can start to see how they can be utilized to advance businesses and society today, and what could be next in the world of AI.

Robotic process automation vs. machine learning – the beginning

The term ‘artificial intelligence’ was first coined by American computer scientist John McCarthy in 1956, but the concept actually developed six years earlier. British computer scientist Alan Turing created a machine much earlier – during the second world war – that could converse with humans, mimicking intelligence indistinguishable from that of a human. Furthermore, Turing suggested that machines would be able to think, and thus be indistinguishable from humans by the year 2000. Side note: the ‘Turing Test’ is still used today to test AI technologies. He argued that if you could have an in-depth conversation with a machine and could not distinguish that it was a machine, then you have proven the existence of real thinking AI.

In the decades following Turing’s first discoveries, scientists became increasingly serious about research on the promising technologies of AI. Soon, programmers and researchers started on specific capabilities that evolved into subsets of AI. One of the first of these was what we call machine learning (ML).

The birth of machine learning in artificial intelligence

Machine learning was one of the first true developments of artificial intelligence. In the 1950s, researchers realized that much of human “intelligence” is based on probabilities – predictions based on past experiences and current inputs. For example, when you see a stoplight turn yellow, you use your knowledge of distance plus your current speed plus past experiences in your car to make a prediction about whether to stop or continue through the light. In theory, computers should be good at this, at making models and predictions based on probabilities.

In 1959, Arthur Samuel, an employee at IBM, used this concept to create the first machine learning algorithm – a program that could play checkers. He used training datasets from past games to “teach” the program a scoring algorithm based on the current board to identify the best possible next move. His program also recorded and incorporated data from the games it played to continuously improve its performance – and machine learning was born.

From there, machine learning became the field of study that gives computers the ability to learn on their own without being explicitly programmed. Machine learning programs can learn from training data as well as from experience to improve over time. Over the next several decades, researchers and theorists continued to develop ML programs that used data inputs to create mathematical, probabilistic, and statistical models. However, computing limitations kept machine learning from truly making an impact in commercial settings until more recently.

The birth of robotic process automation in artificial intelligence

Robotic process automation, on the other hand, came much later than machine learning – it wasn’t developed until the early 2000s. One of the reasons that it was developed much later than ML is its purpose: RPA solutions automate rule-based, repetitive processes. So, until personal computers and data entry became integral parts of operations, RPA wasn’t needed.

From a technology standpoint, there were also two key developments in the 1990s that led to RPA: First, screen scraping software could now extract data from websites, programs, and documents, and second, workflow automation tools were developed. RPA still uses these basic technologies, but works on the user interface instead of relying on APIs. RPA bots can also work around the clock without human intervention to complete many rule-based tasks that require pulling data from one program or interface and using it in another.

Financial services firms were one of the first to deploy RPA in a business setting, using the software to facilitate business processes. Its value was recognized quickly across industries, and RPA solutions have exploded in popularity as companies realize it is the perfect way to automate many of their most burdensome manual, costly repetitive tasks. Walmart reportedly uses RPA to answer employee questions and retrieve pertinent information from audit documents. American Express Global Business Travel uses RPA to automate airline ticket cancellations and refunds1. As more and more people start to understand the potential use cases of RPA, the more aspects of our digital life are managed through RPA technology.

RPA & machine learning applications today

Artificial intelligence, including machine learning and RPA technologies, are only as powerful as the computers and the data they work from. From the 1950s until 1990s, computer scientists found it difficult to be truly successful with developing the full capabilities of artificial intelligence – until the processing power of computers were more advanced and capable of harvesting enormous amounts of data. The immense computing power, speed, and data available today have created perfect opportunities to effectively deploy these solutions. Now, almost every industry is using RPA and ML to create efficiencies, increase capacity, and improve the customer experience.

Companies employ RPA solutions to streamline administrative tasks, from customer service to e-commerce deliveries to payroll. Machine learning is improving business processes, such as credit card companies that automatically place holds on cards with suspicious transactions or marketing teams that autonomously test and optimize ads and product placements.

For example, think about a typical e-commerce order: machine learning algorithms predict what products would appeal to you based on other customer’s buying behavior, and then use testing to optimize advertisements to you. Once you place the order, RPA takes over to send order confirmations, update internal inventory, notify the warehouse, and then emails you a shipping update. If you purchased additional items, that information was stored by the machine learning algorithm to improve advertising for other customers.

However, even though RPA and ML are being used with increasing frequency, they are still often siloed and limited to one specific use case. Some innovators are working to change that. Machine learning that shares globally and learns from experience is what will drive the disruptions that shape our future – think voice assistants and self-driving cars. And while it’s exciting to picture a future where we tell our car where we want to go and it takes us there while we watch TV in the backseat, it is the applications of ML and RPA in healthcare that will truly change our lives for the better.

Machine learning vs. RPA in healthcare

From the beginning of artificial intelligence development, researchers wanted to find practical applications of the technology, specifically in healthcare and medicine. After all, improved customer service may be nice, but it isn’t exactly life-changing. The first healthcare application of artificial intelligence was a program called MYCIN, created in the 1970s at Stanford University. MYCIN used physician input and a coded rule-based system to diagnose and recommend treatment for blood infections, however, the program was never used in practice because data collection and personal computers were not part of current medical practice.

Just like in other industries, it took a long time for practical applications of artificial intelligence to develop in healthcare due to lack of data and computing power. That’s why the first real use of machine learning was in medical research, not in the hospitals. For example, the University of Chicago developed a program in 1999 that read mammograms and detected breast cancer 52% more accurately than radiologists2.

RPA was not used in healthcare until even more recently. After being adopted by other industries to automate many enterprise processes, it was realized that these same technologies could be applied to the back office of healthcare. In fact, with the vast amounts of data gathered and required by EHRs, healthcare is a perfect candidate for RPA. Now, many health systems are using RPA to address the administrative workflows that are bogging down their employees.
Today, it is still sometimes seen as ML versus RPA solutions: which one is right for my hospital, which one has the fastest ROI, which one will solve my most critical problems? But as ML, RPA, and other artificial intelligence technologies develop, it is the combination of the technologies that will be the most powerful. Each day we are getting closer to Turing’s vision of AI that mimics human reasoning. By combining RPA and ML, along with other AI technologies such as CV, neural networks, deep learning, and more, these historically discrete solutions are working together to solve more problems and automate more complex tasks than ever before.

Instead of working in a silo, machine learning can enable RPA bots to be smarter and make decisions. And RPA bots can help machine learning programs constantly and autonomously take in data to learn from, creating optimal models. In fact, with the vast amounts of data now being generated and collected across platforms, RPA plus ML may be the only way to uncover the insights hidden in today’s big data. Instead of the static RPA bots of the past, the future model may be an AI-enabled workforce that can get smarter over time, make decisions, and automate more complex tasks. This new generation of smart bots will be able to do everything from predicting readmission risks to developing new, more affordable drugs and vaccines.

AI workforces will save healthcare

AI in healthcare holds particular promise: in an industry rife with problems – physician burnout, a nursing shortage, and rising costs, to name but a few – healthcare RPA and ML may be able to answer our biggest challenges. In the future, RPA and ML will work together to maximize the knowledge of the data being collected and reduce the administrative burden on our healthcare employees. Armed with this knowledge and time, our healthcare system will be better equipped to improve the health and care of every patient.

At Olive, we are working every day towards this future. We truly believe that an AI-powered workforce may be the answer to the problems our current system is facing, and will unlock the future we imagine by overcoming the limits of human capacity.


  1. https://www.cio.com/article/3236451/what-is-rpa-robotic-process-automation-explained.html
  2. https://www.cancernetwork.com/articles/computer-technology-helps-radiologists-spot-overlooked-small-breast-cancers

Investing for Enterprise Impact Whitepaper

A Healthcare Executive's Guide to AI Workforce Models

Read this white paper to understand the total cost of ownership of various purchasing models for AI and automation, so you can chart the right course as you invest.

Get the White Paper

Related Articles

Learn more