Contact Center LLM Context Enrichment for Agent Interaction History
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Contact center agents often lack awareness of a user's past interactions and recent events, leading to inefficient processing and resource wastage due to repeated information requests and limited contextual information in legacy systems.
Innovation Solution
Implementing a generative large language model (LLM) AI system that enriches raw event data with semantic context using CIX cards, providing agents with contextualized event information and dynamic updates during interaction legs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If agents rely on legacy systems with limited contextual information, then system complexity is reduced, but agent effort and processing time increase due to repeated information requests
Solution Approach 1:
The patent introduces an intermediary component (contextual information system/AI assistant) that sits between the legacy event data storage and the agent interface. This intermediary enriches raw event data with semantic context and retrieves relevant historical interaction information, presenting consolidated contextual information to agents without requiring them to access multiple legacy systems directly, thus reducing agent effort while managing system complexity through modular architecture
Solution Approach 2:
The system performs preliminary actions by pre-enriching event data with semantic context and pre-retrieving relevant historical interaction information before agents need it. Contextual information is prepared and staged in advance, so when agents interact with the system, the information is already organized and ready for consumption, reducing their effort and processing time
2Reliability
If agents are provided with comprehensive contextual information from multiple sources, then user interaction quality improves, but computing resources and processing time increase
Solution Approach 1:
The system applies partial action by selectively retrieving and enriching only the specific event data and historical interaction information that is relevant to the current user interaction context, rather than processing or retrieving all available data. This targeted approach provides sufficient contextual information for quality interactions while avoiding the computational overhead of processing excessive data
3Loss of information
If the system retrieves and enriches event data in real-time during interaction legs, then information freshness improves, but processing time and agent wait time increase
Solution Approach 1:
The system performs preliminary actions by pre-fetching and caching event data and historical interaction information before interaction legs begin or during idle periods. This allows the system to have contextual information ready in advance, so when agents need it during active interactions, the information is already prepared and can be delivered quickly, maintaining both freshness and minimizing wait time
Solution Approach 2:
The system maintains continuity of useful action by continuously updating and maintaining the contextual information repository in the background during interaction legs. Event data enrichment and historical information retrieval operations continue asynchronously without blocking agent workflows, ensuring information remains fresh while agents work uninterrupted
Data Source
AI summary
The present disclosure generally relates to systems, software, and computer-implemented methods for using generative artificial intelligence to improve user interactions. One example method includes receiving a notification from a contact center application that user interaction events have been generated during an interaction session. Event descriptions for events generated in the session are located in a contact center application use case definition. Event descriptions are enhanced with event information for to generate contextualized event information. The contextualized event information to is added to a generative large language model artificial intelligence context that is provided to a generative large language model artificial intelligence engine. A query is provided to the generative large language model artificial intelligence engine. A query response is received from the generative large language model artificial intelligence engine and the query response is used in the interaction session.