Predictive Response Model for Customer Service Chat Agents

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing instant messaging systems for customer service chats often result in delayed or unsatisfactory responses, leading to customer frustration due to inefficiencies in agent response times and lack of personalized interaction.

Innovation Solution

A predictive model is trained based on attributes associated with the customer service agent, such as enterprise characteristics and communication logs, to anticipate conversational phrases, providing suggested responses to the agent in real-time, thereby improving response speed and relevance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If a predictive model is implemented to provide suggested responses in real-time, then response speed is improved, but system complexity increases

Engineering Contradiction:
Improveresponse speedVSAvoidsystem complexity
Core Design Contradiction:
SpeedVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by training a predictive model on historical communication logs and agent attributes before real-time chat interactions. The model pre-processes and stores predicted conversational phrases that can be quickly retrieved during actual customer service interactions, enabling fast response suggestions without complex real-time computation.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The predictive model acts as an intermediary between the customer service agent and the customer. It receives the current chat context and agent attributes as input, processes this information through the trained model, and provides suggested responses that mediate the interaction. This intermediary layer simplifies the agent's task while maintaining system manageability.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If personalized interaction is provided using agent attributes, then customer satisfaction is improved, but data processing requirements increase

Engineering Contradiction:
Improvecustomer satisfactionVSAvoiddata processing requirements
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The system applies local quality by using specific attributes of the customer service agent (such as expertise level, communication style, and performance metrics) to personalize the predicted responses. Instead of processing all possible data uniformly, the model focuses on relevant local characteristics of each agent to generate appropriately personalized suggestions, reducing overall data processing requirements while maintaining high customer satisfaction.

Inventive Principle:
Principle #3Local quality

3Measurement precision

If contextually relevant phrase predictions are provided, then interaction quality is improved, but model training complexity increases

Engineering Contradiction:
Improveinteraction qualityVSAvoidmodel training complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The model training process is segmented into distinct phases: first, the model is trained on historical communication logs to learn general conversational patterns; second, it is fine-tuned using agent-specific attributes to capture personalized communication styles; third, the model is evaluated and optimized for contextual relevance. This segmentation of the training process manages complexity while achieving high interaction quality through contextually accurate predictions.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3881317B1System and method for accelerating user agent chats
Publication Date: 2024.06.19 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3881317B1 patent drawingFigure 1
  • EP3881317B1 patent drawingFigure 2
  • EP3881317B1 patent drawingFigure 3

AI summary

A method, computer program product, and computer system for identifying, by a computing device, a model for predicting conversational phrases for a communication between at least a first user and a second user. The model may be trained based upon, at least in part, an attribute associated with the second user. At least one conversational phrase may be predicted for the communication between the first user and the second user. The at least one conversational phrase may be provided to the second user as an optional phrase to be sent to the first user.