Robotic Process Automation (RPA) is one of the fastest growing software technologies and one of the most affordable implementation of artificial intelligence. Many enterprises expect the RPA to allow a significant number of employees to complete repetitive data processing tasks quickly, at lower cost and with fewer errors, and give their time and efforts to more important work.
How it works? The simple example is working with your email. Every time you look into your mail, you repeat about the same thing. First, you classify the mail. Junk goes to the trash bin, bills and payment reminders to the “Payments” folder, requests from customers in the “Important” folder, and so on. Then we extract the information from the messages and act in accordance: the accounts are registered and transferred to the accounting department (marked urgently if the payment date is tomorrow), requests are transmitted to the appropriate managers, etc.
The operations of many companies look exactly the same - they implement their information processing scenarios, usually with fairly large amounts of data. Arrays of data elated to interaction with customers through various channels, or information on stocks, deliveries, equipment operation parameters, geolocation of cargos and vehicles - all this is classified and processed based on the rules, and then the information is sent to the appropriate group within the organization for further actions. It is these processes that RPA helps automate. What people used to do in a certain user interface with their hands is now done automatically.
An example with parsing mail is quite vital. American Fidelity Insurance Company is already saving time using RPA solutions. Previously, managers classified and redirected incoming customer messages manually by reading each message, and using RPA and machine learning, it was possible to reduce the processing speed from 45 hours to impressive 3 seconds.
Unstructured data and artificial intelligence
Very often, data processing tasks are rule based. It is not at all difficult to develop a bot that can cope with a predefined structured input. But if the data is not structured, if the rules are not clear and unambiguous, it is required to use artificial intelligence tools to classify and extract information from such sources. For example, machine learning or natural language processing (NLP).
Machine learning will require a fair amount of unstructured content. For example, if you want to automatically extract customer complaints from e-mail, the machine learning model should be trained on a large number of complaints from previous years. However, today there are already pre-trained models that can solve problems right out of the box.
It was claims processing that the American insurance company Safe-Guard Products automated by using RPA. The time required to process a claim was reduced by 75%, and the overall team productivity increased by 30%.
Integration with data mining tools
RPA integrates seamlessly with other tools to extract data without human intervention. A great example is applications that use optical character recognition (OCR) to scan documents. Such applications can work in conjunction with an RPA, classifying and routing structured or unstructured data.
Another example is interactive voice response (IVR) and client request routing applications. RPA systems enable IVR applications to implement both rule-based logic and machine learning algorithms trained in customer service functions.
RPA limitations
RPA does not completely replace people. This technology is not intended to automate complex processes end to end, RPA can be effective in replacing manual labor for repetitive tasks, when input data is more or less predictable, but interaction with other systems is often required to complete processes.
For example, you need to open an order received by e-mail, copy certain fields and paste them into a spreadsheet, and then upload to a web form. This is an ideal job for an RPA, but it is only part of the order process. The work of the financial department for billing and payment control, related operations in ERP and CRM systems - all this covers different departments and requires the interaction of employees. This goes beyond user-interface interactions and requires RPA integration into the broader Business Process Management (BPM) and IT Process Automation (ITPA) platforms, which span enterprise-wide processes and offer integration with a variety of applications.
In 2020, the RPA market will be $ 1.5 billion (Forrester forecast). Today, among the leaders in this segment are UiPath, Pega, Automation Anywhere and Blue Prism, which are at the forefront.
If you add RPA and related machine learning technologies to your application stack and you can creatively apply RPA and artificial intelligence to your tasks, the range of problems that these technologies can solve can be very impressive.
