Use of AI in predictive routing
Genesys predictive routing uses artificial intelligence to route the interactions to the best agent available for the KPI you set. The agent score derived by predictive routing is based on a number of drivers such as level of data available, skill level of agents, availability of agents, timeout period.
What data does Genesys AI use to make routing decisions?
Genesys data models rely heavily on data that is both populated and calculated/derived from other data in the system. The availability of all the required data ensures that the models function at their optimum levels. The following are examples of the features and data sources that predictive routing uses for agent scoring:
- Agent profile data such as skills, tenure, department, certificates, employee type. For more information, see Data requirements for predictive routing.
- Agent performance data such as historic average handle time for a queue.
- Customer history data such as the number of times they called the contact center in the last 30 days.
Currently, the primary sources of data are agent directory (for agents) and analytics (for customer and interaction data).
How are data models created and maintained?
To keep up with changing levels of agent proficiency and customer interaction contexts, the data models continually retrain and learn from the latest features. Genesys Cloud updates the features used for agent scoring with daily data and retrains the data models weekly. No data is retained in models for more than 90 days. This ensures that the latest operational data is taken into account for forthcoming routing decisions.
As newer data models emerge with updated features, Genesys retires the older data models that are no longer relevant. Where models are unused (for example, where predictive routing is deactivated on a queue), they are purged from the system automatically.
Genesys Cloud platform creates and maintains the data models used in predictive routing. For more information on how data is used in agent identification, see How AI model scores agents for predictive routing.
How can we interpret the working of the features and the data model?
Predictive routing uses white-box models that allow gaining insights into how the features contribute to a prediction. Genesys helps users deduce the prediction by presenting a global interpretation that describes the average behaviour of a model. Each input feature is given a percentage/score that represents its ‘importance.’ A high value means the feature will have a larger effect on the model’s predictions and ranking of agents. While a small value is given to ‘unimportant’ features whose contribution is mostly ignored for the model’s predictions. To see the features specific to a queue, follow the documentation available on How AI model scores agents for predictive routing.
Genesys does not use and will never use PII for the agent scoring process. Genesys Cloud only uses transaction conversation data to train machine learning models. The agent profile and the performance data that are used in the scoring process do not contain any agent PII.
How does Genesys ensure that no discrimination is introduced in the agent scoring process?
Genesys does not and will never create data models that require data such as gender and nationality that have the potential to introduce discrimination. To train the data model, predictive routing generates features only using transactional conversation data that do not contain PII. The absence of PII ensures that there is no scope for discrimination in the scoring process.
It must be noted that in order to keep the predictions as accurate as possible, predictive routing uses the most recent historical conversations. This may introduce what is known as a temporal bias.