Insurance agency employees spend a lot of time every day just opening letters and reading e-mails, faxes and other such communication. Each piece of mail needs to be examined in order to determine to which department or employee it should be forwarded. This process is not only time-consuming, but also prone to delays, such as when an employee is ill, or during times when the need to deal with a sudden rush of inquiries arises- for example after a storm. In addition, the amount of incoming mail is constantly growing. What is needed are systems that can efficiently handle this task, lighten the load of the employees, and optimize and accelerate the distribution process.
Automation is possible
Due to the steadily increasing amount of daily incoming mail, this matter is becoming more and more relevant. The question is how this process can be optimized and automated, so that firstly customer service is improved, and secondly the departments and employees are freed up to do their jobs. The good news: not only do potential answers to this problem exist, but fully designed and ready-to-use solutions are already available. Nevertheless, only a few companies have considered this issue at all. The solution is called automated incoming mail classification.
But how can a computer program sort documents? How can it know which letter should go to which employee? Intelligent incoming mail classification systems achieve this goal by analyzing documents semantically, meaning that they understand the contents and in a broader sense can use predictive analytics to determine how the document should be classified.
Stated simply, the term “predictive analytics” describes learning from the past for the future. Personal experience is to humans what predictive analytics is to software. The software recognizes certain patterns in these "experiences" and uses them to reach conclusions about future behavior. Using this automatic classification, each letter is forwarded directly to the relevant department. There, the documents can be immediately processed and dealt with. This optimizes the entire handling process. Another advantage of this system is that it learns from its mistakes. If a documents was incorrectly classified and then manually corrected, the system remembers this. The longer the system is in use, the more information the application collects and the more accurate the classification becomes.
Structured and unstructured
To an intelligent incoming mail classification system, it doesn’t matter whether the data that arrives at the company is structured, e.g. a completed online form, or unstructured, for instance in the form of an e-mail text. The system treats both possibilities equally. Flexibility is one of the strengths of an intelligent system. Whether an e-mail, a scanned letter or the now ubiquitous social media post - every piece of written information is analyzed, sorted and classified. This also applies to future input channels, such as the Internet of Things. The role of “big data”, particularly in the insurance sector, is gaining increasing importance.
The initial configuration of automated incoming mail classification can be done with very little effort. After the integration into the company's own IT, the data sources can be linked by means of connectors to a broad variety of different data sources. Of course, these also include typical data sources such as network drives, Microsoft SharePoint, and a variety of ECM systems. Then the training begins. In training mode, the systems learn to select already classified documents based on existing criteria. Thus, the system becomes more intelligent with each document it processes. The more pre-classified documents from recent days, weeks or months are used for this training, the higher the learning success rate is. The learning process itself requires only a few milliseconds per document. Should the system make an error, the manually corrected document is re-submitted to the system, which then saves the correct classification information.
One challenge for automated incoming mail classification is the variety of input items. Letters on paper, e-mails, often with attachments that consist of several pages, faxes and even postings in social media channels – the system has to be able to process and analyze all of these sources. The selection criteria used by the system to determine which department should receive which document are decisive for the success of the system. Mindbreeze InSpire uses around 3000 different characteristics to classify each document. The solution is already operating in major insurance companies. Because the system is self-learning, additional selection criteria can be added at any given time. Considering the fact that even greater amounts of data are going to have to be processed in an even shorter amount of time in order to satisfy customer demands in the future, there is no getting past the need for automation in this field.