AI Visual Content Filtering for Contact Center Agent Protection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contact center agents are exposed to prolonged negative engagement content, such as angry threats or insults, which affects their mental health and job performance, and existing de-escalation methods are often ineffective.
Innovation Solution
Implementing AI-driven content filtering systems that analyze audio and visual content in contact center engagements to determine negative emotional states and replace the content with filtered versions, using AI models trained for sentiment analysis and relevance scoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If contact center agents directly engage with users experiencing negative emotions, then customer service completeness is maintained, but agent mental health deteriorates
Solution Approach 1:
The patent introduces an intermediary system comprising AI models and content filtering software that sits between the user and the agent. This intermediary analyzes user content in real-time, determines negative emotional states, and filters or modifies the content before presenting it to the agent, thereby protecting the agent from harmful negative engagement while maintaining service continuity
Solution Approach 2:
The patent extracts and removes the harmful negative content from the engagement flow. The content filtering system identifies negative emotional content, separates it from the legitimate communication needs, and either filters it out or presents a modified version to the agent, allowing the agent to serve the customer without direct exposure to harmful elements
2Reliability
If content filtering systems are implemented to protect agents, then agent wellbeing is improved, but system complexity increases
Solution Approach 1:
The patent creates a universal content filtering system that handles multiple types of content (audio, text, video) through a single integrated architecture. The AI models and filtering software are designed to work across different communication modalities and contact center platforms, reducing overall system complexity through standardization while providing comprehensive protection
Solution Approach 2:
The system employs AI models that automatically analyze content, determine negative emotional states, and make filtering decisions without requiring manual intervention. The self-service nature of the AI-driven filtering reduces operational complexity and allows the system to scale without proportionally increasing management overhead
3Object-affected harmful factors
If AI models analyze content in real-time to filter negative engagement, then negative content exposure is reduced, but processing time increases
Solution Approach 1:
The patent implements preliminary analysis mechanisms where the AI models are pre-trained on extensive datasets of negative emotional content. This preliminary training enables the models to quickly recognize patterns and make rapid filtering decisions during real-time engagement, minimizing processing time while maintaining accurate negative content detection
Solution Approach 2:
The system employs periodic sampling and threshold-based filtering where not every single content element requires full AI analysis. Instead, the system uses a combination of quick heuristic checks and periodic deep AI analysis, only applying full computational resources when negative content is suspected, thereby reducing overall processing time while maintaining protection effectiveness
Data Source
AI summary
Visual content filtering is performed to replace visual content initially presented to a contact center agent device from a contact center user device during a contact center engagement. A determination is made, at a first device of a contact center agent, to filter visual content of a video stream of the contact center agent for a contact center engagement with a contact center user. Filtered content corresponding to the determination to filter the visual content is then obtained at the first device. An updated video stream is generated at the first device by replacing the visual content with the filtered content. The updated video stream is then output in place of the video stream for rendering at a second device of the contact center user during the contact center engagement.