Contact Center Agent Routing Using Dynamic ML Feature Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional contact center agent selection methods rely on static configurations that fail to consider dynamic parameters such as agent skill, historical interaction data, customer emotions, and supervisor requirements, leading to inadequate performance in complex scenarios.
Innovation Solution
Implementing a machine learning model that dynamically selects contact center agents based on a growing set of attributes from incoming communications, using supervised learning to adapt to changes in communication patterns and continuously update agent selection criteria.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If static configuration of routing rules is used, then implementation simplicity is maintained, but routing accuracy and adaptability deteriorate when multiple dynamic parameters are considered
Solution Approach 1:
The patent transforms the static routing configuration into a dynamic machine learning model that continuously learns from historical data and adapts to changing patterns. The system dynamically adjusts routing decisions based on real-time parameters such as customer emotions, agent performance, and interaction outcomes, replacing fixed rules with adaptive algorithms that evolve over time.
Solution Approach 2:
The patent replaces the mechanical system of static routing rules with an intelligent system based on machine learning models. Instead of manually configured decision trees and if-then rules, the system uses trained models that automatically process multiple parameters and generate routing decisions, substituting rigid mechanical logic with flexible computational intelligence.
2Measurement precision
If multiple dynamic parameters are considered for routing, then routing accuracy improves, but system complexity and computational requirements increase
Solution Approach 1:
The patent merges multiple separate parameters (agent skills, customer emotions, historical performance, interaction outcomes) into a unified machine learning model. Instead of processing each parameter through separate rules, the model integrates all inputs simultaneously, capturing complex interactions between parameters that would be difficult to encode in individual rules.
Solution Approach 2:
The patent transforms static parameter thresholds into dynamic learning targets. The system continuously updates parameter weights and relationships based on historical data, allowing the importance of each parameter to change over time based on what actually correlates with successful outcomes, rather than relying on fixed predetermined thresholds.
3Productivity
If static routing rules are used, then system scalability is limited, but implementation cost is reduced
Solution Approach 1:
The patent implements a self-learning system that automatically improves its own performance without manual intervention. The machine learning models continuously train on new data, automatically adjusting their parameters and decisions based on observed outcomes. This self-service capability allows the system to handle new scenarios and patterns that emerge over time without requiring manual rule updates.
Solution Approach 2:
The patent incorporates feedback loops where routing outcomes are continuously monitored and fed back into the training data. Successful routings reinforce certain patterns, while unsuccessful ones trigger model retraining. This feedback mechanism enables the system to learn from experience and progressively improve routing efficiency across diverse scenarios.
Data Source
AI summary
In one example embodiment, one or more machine learning models of at least one processor determine an agent of a communication center to receive a communication from a user according to one or more objectives that optimize a corresponding metric from interaction of the user and the agent. Features for training the one or more machine learning models are dynamically selected based on relevance of the features to attaining the one or more objectives. The at least one processor routes the communication to the agent of the communication center.