Applying NLP for case classification

This demo shows you how to enable Case classification with AI and test it in a real-world scenario.

Consider a scenario that involves an insurance company, U+ Insurance. This company is currently handling a high volume of car insurance claims. The process they follow is traditional and manual; experts review all Cases and manually categorize and route them based on the description provided by customers. This process often results in delays and errors due to the sheer volume of Cases and the human element involved.

Recognizing the need for a more efficient and accurate process, the company decides to utilize natural language processing in Pega Process AI.

The goal is to automate the categorization and Case routing, which reduces the time it takes to process claims and minimizes the possibility of errors as a result.

An Application Developer must enable the AI-powered Case classification prediction. If the desired category field is nonexistent in the application, the developer begins by creating it in the Case and defining the topics. The system uses the choices set by the Application Developer as topics for the new Accident Category Prediction. This is a crucial step as it sets the foundation for the AI to understand and categorize the Cases. Now, the Application Developer must enable AI and specify the Case description provided by the customer as the input text for AI analysis.

After the Application Developer enables AI, the system automatically creates a new Prediction in Prediction Studio. When insurance claims arrive, the Prediction uses AI models to recognize the topic of each claim and categorize it based on the Case description provided by the customer. The topics correspond with the choices set in the new Accident Category field.

At first, the models are not trained. Data Scientist trains and builds the models in Prediction Studio. Data Scientist uses a provided training data file, which contains examples of Case descriptions and their corresponding topics. The topics match accident categories that the prediction needs to detect.

After the models train on the data, the text Prediction undergoes testing in Prediction Studio. The Data Scientist confirms that the AI correctly detects the topic of each message by providing the AI with sample Case descriptions and checking if the AI correctly identifies the accident category.

After the models complete training and testing, the Application Developer steps back in to validate the functionality of the automatic AI accident category prediction and Case routing in the Process AI Example Application. The developer creates a new Claims Case for each Persona and validates the detected category. This step helps ensure that the AI is accurately categorizing and routing the Cases in a real-world scenario.

Finally, the Application Developer verifies the pre-configured intelligent Case routing. The system automatically routes the Cases to the correct Work Queue based on the predicted accident category. This action eliminates the need for manual classification and ensures that the appropriate teams handle Cases, which improves the efficiency of the claims processing workflow.

This demo has concluded. What did it show you?

* How to enable AI Case classification in Pega Process AI by creating a new accident category field and defining topics.

* How to train and test models in Prediction Studio.

* How to validate the automatic AI accident category prediction and Case routing by creating a new Claims Case for a test Persona.

* How to verify the effectiveness of intelligent Case routing, which automatically routes Cases to the correct Work Queue based on the predicted accident category.