Jump to: navigation, search

Quick Start

To use Genesys Predictive Routing (GPR), you'll need to install and configure the following products and components:


  1. Install any Genesys components that aren't already part of your environment. You'll need:
  2. Install Genesys Predictive Routing, which consists of the following components:
    • Agent State Connector (ASC)—Synchronizes agent state data from Stat Server (agent availability) and Configuration Server (agent profile) for use by AICS.
      In ASC release and later, you can choose to have ASC read agent availability information from Universal Routing Server (URS) rather than from Stat Server, minimizing the number of components ASC connects to.
    • Integrate with Genesys Routing
    • AI Core Services (AICS)—Provides the Genesys Predictive Routing scoring engine, the user interface, and the API.


To complete your setup of Predictive Routing, configure the following components:

  1. Set the desired values for the configuration options.
    You use configuration options to configure a wide range of application behavior, including:
    • The mode GPR is running in, which might be off
    • Login parameters and access URLs
    • KPI criteria to decide what makes for a better match
    • Scoring thresholds, agent hold-out, and dynamic interaction priority.
    • Many other important functions
  2. To configure how the match between interactions and agents is determined, configure Predictors and Models, as explained in the Predictive Routing Help.

Import Data

Using the Predictive Routing interface, you import a dataset that is available in CSV format. A dataset is a collection of raw event data. The primary purpose of a Dataset is to be the source of Predictor data.

  • GPR automatically analyzes the data and creates a schema, identifying the various types of data you are importing.
  • You can adjust the schema during the import process.
  • After the Dataset has been imported, you can append additional data as long as it is consistent with the schema that has already been established.

Create Predictors and Models

Predictors are based on the dataset information that you have imported and that has been analyzed into a schema.

  • A Predictor defines a view on that underlying dataset. It can select from some or all of the data in the dataset; you can use a predictor with multiple datasets.
  • A Model is based on a Predictor, and uses the same target metric or KVP as that Predictor. You can configure multiple Models for each Predictor. These Models can use different selections of the features available in the underlying dataset. Models are the objects actually used to perform agent scoring and interaction matchups.

As you configure a Predictor, you can choose which metric you want to work with, what kinds of situations you want to evaluate, and other parameters, constructing a way to determine the Next Best Action in the specified situation, based on the possible actions available at that time. As you gather more data, you can add that new data to your dataset, and have the Predictor test against the actual results coming in, enabling you to refine how successful your Predictor is.

Reporting and Analysis

You can report on various parameters, such as:

  • The success of your predictors.
  • The results of A/B testing.
  • The factors affecting a KPI you are trying to influence.

Reports are available through the following reporting applications:

This page was last edited on July 19, 2019, at 19:53.
Comments or questions about this documentation? Contact us for support!