At RoboRana, we are constantly challenging the boundaries of RPA and exploring new possibilities to improve our intelligent automated solutions. We strongly believe that the combination of RPA with cognitive technologies based on Artificial Intelligence (AI) will enhance and extend the possibilities of the Robotics Process Automation Journey at our clients. To prove the power of cognitive RPA, we created a few months ago some promising and insightful demo’s in which we combined cognitive technologies like chatbots and intelligent Optical Character Recognition (iOCR) with RPA.
However, we noticed at our clients that there are many manual business processes that handle unstructured information or data inputs directly coming from end-customers. For example, a request for a new credit card or an inquiry about an address change are now being emailed to a generic mailbox that is constantly being monitored by the back-office. Next, the mails are manually interpreted by the administrator who then starts the manual execution of the dedicated process (new credit card or address change in this case).
As these processes can be easily automated by our RPA-bots, we decided it was time to have a look at Natural Language Processing (NLP), and investigate how this smart technology could assist our customers and their business processes. We teamed up with our friends from Craftworkz to create a small proof-of-concept in which we receive various unstructured e-mail requests and analyze them with NLP to understand the correct action to be taken and to retrieve the details of the requests.
The process begins by reading a dedicated mailbox that gets filled with various kinds of unstructered requests. It is difficult to filter out these mails by just using mailbox rules, so we created a process that our RPA-bot reads the email body, sends it to the NLP service (Python) for analysis and then executes the specific process according the action and information extracted by NLP.
The NLP service extracts the correct action from the text (address change, new credit card, …) and subsequently analyzes the text to find all the necessary information, such as the new address and client number. Once the unstructured text has been analyzed, the NLP service communicates this information back to the RPA-bot, which can then continue the execution of the task in a correct manner. This can be by either adding the item to the correct work queue, or immediately handle the complete request.
Although, we only have used a small aspect of Natural Language Processing in this demo, it is clear that this technology lends itself perfectly to be used in combination with RPA. Apart from detecting the required action from a text, it can also be used to analyze the sentiment in the text or request. This sentiment output can be then used to prioritize certain requests.
Besides the fun of creating this demo, we have also learned how to quickly and efficiently integrate our RPA processes with other cognitive services or technologies which is essential for our road to intelligent automation.