Robotic Process Automation Enhanced by Artificial Intelligence
Jul 15, 2019
For the past few months, CTG Luxembourg has been working on the development of an RPA solution supported by artificial intelligence. Able to make decisions and carry out its tasks independently, this tool has everything to seduce companies in the midst of digital transformation.
RPA or "Robotic Process Automation" is becoming more and more popular among companies that need to accelerate their digital transformation today. RPA solutions provide the considerable advantage of automating a series of repetitive actions, whether they are business or purely IT processes. These low value-added tasks are nowadays performed by employees. Thanks to RPA solution, they can refocus on more complex work, which brings added value for both employees and employers.
A Robotic Process Automation that learns
The problem is that conventional RPA needs to be very framed to work optimally. "These robots expect to receive highly-structured data. They must also have integrated all possible scenarios in advance to be able to operate properly," explains Mr. Frédéric Warrant, Solution Architect at CTG Luxembourg. "This means in particular that they cannot make decisions autonomously. To overcome this pitfall, we had the idea of combining the RPA tool provided by our partner UIPath with an in-house developed ‘machine learning’ technology."
In a few months, CTG Luxembourg's teams have therefore developed a solution that can integrate and analyze a large number of structured and unstructured data (using "Natural Language Processing" technology) but also - in the long term - make decisions and launch procedures based on this information. "In practice, for example, our solution can be used from an email list. The e-mails received by a company are already categorized and lead to the opening of a ticket, which leads to the launch of a procedure. All this information can be integrated into our solution, which will teach what is the right procedure to launch in this or that case. When a new e-mail arrives in the system, the RPA, supported by artificial intelligence, will then be able to launch the appropriate procedure itself, make the right decision without the help of a collaborator," illustrates Mr. Frédéric Warrant. “In addition, in the learning process of artificial intelligence, human input remains necessary.”
Many possible applications
If this combination of RPA and artificial intelligence can work perfectly with text data, this may also be the case for voice or images. "In the same way as with the text, the system we have developed could analyse images, categorize them, and launch appropriate procedures. The voice would be transformed into text before being analysed. In addition, the solution could also detect intonation variations and consider them in its decision, says Mr. Frédéric Warrant. We're not there yet, but these are developments that are of the realms of the possibility."
However, to work properly, this solution must be able to rely on quality data (i.e. correctly encoded).
"In addition, in the learning process of the artificial intelligence, human input is still necessary," says CTG Luxembourg's Solution Architect. Sometimes, artificial intelligence is not confident enough in the response it must provide to a problem. In this case, we can program the solution in order that the human has to decide. The IA can also make mistakes that will be corrected by a human, especially if it asks the robot to start a procedure that is not appropriate. However, every evening, the IA is regenerated by integrating this new data, which allows it to avoid making the same mistakes or to better know how to react in certain cases."
“With our system, we can increase the business while maintaining attractive margins.”
At the present time, the solution developed by CTG Luxembourg stops when the necessary procedure to solve the task is launched. However, this last step should be soon taken over.
"Our RPA is already tested within our support center. The tool allows us to provide better customer service, being faster and managing more easily the request peaks. Obviously, our wish is to offer it to our clients as soon as it is fully operational," explains Mr. Frédéric Warrant. This solution meets the current challenge of companies, which need to digitize their business to be able to evolve and sign new contracts, but which also need to hire many new employees to obtain it. However, the increase in the total payroll reduces companies' margins. "With our system, on the contrary, we can increase the business while keeping attractive margins," says Mr. Frédéric Warrant.
It remains to be seen in concrete terms how to implement these systems within companies.
For Mr. Frédéric Warrant, an RPA project must be segmented, flexible in its execution, and fast in its implementation. The agile method is, therefore, the reference to follow. It ensures that each process automation is linked to a "sprint", i.e. a rapid repetition of a development process. The team in charge must also be multidisciplinary, composed of technicians (RPA experts), process modeling specialists, and specialists in the profession as well as the client himself. At CTG, we also insist on "governance by operations", i.e. an implementation solution that has to be as close as possible to the operational needs of the profession and that allows us to monitor the evolution of these needs on a regular basis. A way to maintain a high degree of flexibility in the use of this precious auxiliary, which is the RPA.