Technology Blogs by SAP
Learn how to extend and personalize SAP applications. Follow the SAP technology blog for insights into SAP BTP, ABAP, SAP Analytics Cloud, SAP HANA, and more.
Showing results for 
Search instead for 
Did you mean: 
Product and Topic Expert
Product and Topic Expert
Robotic process automation (RPA) is probably this year’s most hyped term in the digital worldAnd that’s for a good reason. RPA is an innovative technology, automating business processes using software robots that perform tedious and repetitive tasks – error susceptibility equals zero.

Imagine the magic!

Thanks to robots, the human employee no longer has to waste time on nerve-wracking, time-consuming, and manual activities. The result is relieved and happy employees with extra time for tasks that really matter, improved work results, adding an extra pinch of value to your employees’ daily hustle.

No wonder RPA has already proven to be an intelligent and efficient driver to push digital transformation in enterprises by increasing productivity, reducing cost, and ultimately leading to revenue and business growth. Hence, RPA is more than just a buzzword, which is also why its market size is expected to hit $3.97 billion by 2025.

For a deep dive into “What is RPA,” have a look at the earlier blogs in the “Intelligent RPA” series.

Next level: making RPA intelligent

So, we learned that digital bots automate repetitive and routine tasks – thereby significantly increasing productivity and employee and customer satisfaction.

On other terms, however, these bots are not able to draw any conclusions from their actions; they are not “intelligent.” This is where machine learning (ML) comes in.

Connecting machine learning with RPA is meaningful whenever business automation is pursued in an integrated and strategic manner. In order to automate business processes efficiently, future-oriented, and strategically, combining these two technologies is essential.

By infusing intelligence into RPA, hence combining machine learning capabilities with process automation, we can design an advanced form of RPA – a bot that is able to analyze, comprehend, and draw conclusions from both structured and unstructured data. This powerful symbiosis is consequently able not just to process, but effectively use data. This newly created intelligent RPA analyzes data before acting on it, continuously learns from data, becomes more intelligent over time, makes smart decisions based on previous learning.

Therefore, automating processes with the help of RPA and ML especially makes sense whenever a huge amount of data needs to be processed, analyzed, compared, and structured. While ML covers the task of thinking and learning, RPA executes. ML functionalities that come to play in connection with RPA are technologies such as image and speech recognition or document information extraction, for example.

A real-life example: automating accounts payable

Let’s take a very common business case. Worldwide, every day, millions of accountants receive invoices from their suppliers that notify them about payments to be made.

Thirty years ago, these documents were printed on paper made out of wood and water, put into envelopes, even delivered by a postal agent.

Today, thanks to the digital transformation, invoices are reaching the accountant as PDF files attached to an email. What an improvement!

The bad news is, the whole process around it remains the same.

  • Processing invoices is time-consuming and a rather dull task for accountants.

  • These highly qualified employees must invest a significant amount of working time to read the invoices and select and extract the relevant information manually, e.g., the supplier name, invoice number, total amount to be paid, payment date, and so on.

  • After manual extracting the data, employees open the accounting module in their organization’s ERP application, enter the data into a form, and click a button to validate the entry.

End of the game. If we may call that a game. The manual processing of an invoice takes a couple of minutes per invoice … for millions of invoices per day worldwide.

So what can simplify this process? Easy – automate as much as possible with the help of intelligent RPA!


  • In the first step of the process, an RPA bot opens the accountant’s inbox and identifies the emails related to invoices with an attached document.

  • The bot takes the email attachment from the accountant’s inbox, reads it, and sends it to an intelligent ML application called accounts payable, where the document is processed and invoice details obtained.

  • The application uses ML capabilities trained to process every kind of invoice, from any kind of company in the world; automatically extract the data; and hand it back to the RPA bot.

  • In the last step, the bot creates the supplier invoice in the system, where all running jobs can be monitored by the employee (due to the RPA bot scheduling them).

End of the game. Yes, for an RPA bot using ML capabilities, this process is a quick and simple game, taking only a few seconds per invoice. And more than that, it is free of human errors.

Now, multiply one minute by the millions of invoices sent daily worldwide, and you can imagine the ROI created by the smart combination of RPA and ML technologies. This combination ultimately frees up employees’ time so they can focus on tasks that really matter.

Bridging the gap between academia and industry

While SAP has several machine use cases like this one, continuous research is driving progress in this field. The SAP ML research team focuses on solving tough ML problems with practical applications across multiple domains, such as text, image, or video. The emphasis lies on generic ML techniques that have the potential to impact a wide range of business use cases across finance, procurement, logistics, and travel management.

Through multiple bilateral collaborations with the world’s best research institutions, SAP Leonardo research enlarges its expertise and actively contributes to the continuous development of ML and deep learning solutions. Once ready to scale, the ML models can turn into new enterprise solutions or empower SAP’s existing ones.

For more information

This article, co-written with ivonac, originally appeared on the SAP Analytics blog.