Pega Process AI for intelligent automation

Prediction Studio is the dedicated workspace where Data Scientists build, monitor, and update AI models to predict various case outcomes. The workspace provides everything needed to author, deploy, govern, monitor, and change predictions. Prediction Studio has five work areas:

• Predictions: Create and manage prediction configurations.
• Models: Build, monitor, and update AI models.
• Data: Manage Data Sets and Data Flows.
• Reports: Monitor model performance and accuracy of Predictions.
• Settings: Configure governance and operational parameters.

The following figure shows the five work areas that are available in the navigation pane of Prediction Studio:

On the Predictions landing page, you create and manage Predictions. There are three types of Predictions, but Process AI focuses on the Case Management Prediction, as shown in the following figure:

Create Predictions by using one of the following outcome templates or a custom one based on business requirements:

• Case completion
  Predict the probability of successful Case completion to help you prioritize Cases that provide the highest potential value.
• Missing SLA
  Predict the probability that a Case will miss its SLA to help you prioritize Cases at risk of delay and maintain compliance.
• Claims fraud
  Detect potential fraudulent claims by analyzing historical patterns and anomalies. Use this template to route suspicious cases for additional review. This template is frequently used in insurance and financial services.
• Custom prediction
  Create a tailored Prediction for unique business scenarios. This template gives you the flexibility to define your own outcome and choose your model type.

Use Case Management Predictions to support decisions in business processes by integrating AI-driven decisions directly into your Case Type. You configure Predictions in the Decision Step in a Case Type. Based on the outcome of the Prediction, the system routes assignments, branch processes, or validate information.

For example, the following Case Type handles incoming car insurance claims:

Use the outcome of the Prediction in the condition of a Decision Step instead of a business Rule. Based on the condition, the system routes a Case to a fraud expert when the Prediction flags the claim as abnormal, as shown in the following figure:

Additionally, you can configure a Prediction widget in the Case view to display Prediction results.

In the following figure, Prediction widget shows fraud or the probability of missing an SLA. By clicking Learn more, you can see the details of the model and whether it received training with adequate data (in this example, the probability of Case completion).

The widget also displays which predictors contribute positively (increasing the predicted value) and which contribute negatively (decreasing the predicted value).

Process AI supports natural language processing (NLP) and text analytics features that enable automated Case classification. For example, with NLP, Process AI can automatically identify the accident category in an insurance claim based on the Case description provided by the customer.NLP reduces manual effort and improves routing accuracy by automatically sending the case to the correct Work Queue.

For more information about configuring Case classification prediction, refer to Applying NLP for Case classification

Decision strategies and event processing are other supporting features from the Decisioning Services category. In Process AI, these features help optimize decision points in a business process. You can run a Decision strategy from the Process flow in a Case Type to make decisions within the context of a given Case. Or you can run a Decision strategy as a part of a real-time Data Flow run to determine the best action based on incoming events.

For example, combine a Decision strategy with event processing to triage alerts and automate decisions, such as when to open an investigative Case and which team should receive the Case.

For relevant training materials about Process AI use cases and configurations, see Pega Process AI Essentials.

To begin using Process AI features in your applications, you require a Pega Platform system that meets the technical and sizing requirements for AI, Decision management, and event processing. For more information, see Provisioning Pega Process AI.

Check your knowledge with the following interaction: