Adaptive Recommendation Engine for Agent Assist Platforms

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing chat interface systems often provide recommendations to agents that are distracting and not relevant, leading to longer communication times and poor user device performance.

Innovation Solution

An agent-assistance platform that analyzes the usage of previously provided recommendations to determine whether to provide subsequent recommendations, suppressing recommendations that are likely to be distracting or ignored by agents.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If recommendations are provided frequently to agents, then the system aims to improve issue resolution, but the recommendations become distracting and lead to longer communication times

Engineering Contradiction:

Improveissue resolution efficiencyVSAvoidcommunication time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary analysis of conversation context and agent behavior patterns before providing recommendations. By predicting when recommendations will be useful based on historical data and current conversation state, the system avoids providing distracting recommendations, thereby reducing communication time while maintaining issue resolution efficiency

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system monitors agent interactions with recommendations and uses this feedback to adjust future recommendation timing and content. By tracking whether agents actually use recommendations and how conversations progress, the system learns to provide recommendations at optimal moments, balancing productivity improvement with minimizing communication time

Inventive Principle:

Principle #23Feedback

2Ease of operation

If the system provides automated recommendations, then support for agents is improved, but system resources are consumed excessively

Engineering Contradiction:

Improveagent support qualityVSAvoidsystem resource usage

Core Design Contradiction:

Ease of operationVSUse of energy by moving object

Solution Approach 1:

Instead of providing recommendations in all possible situations, the system applies partial action by selectively providing recommendations only when prediction models indicate high likelihood of usefulness. This reduces system resource consumption while maintaining adequate support quality for agents

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The system dynamically adjusts recommendation provision based on changing parameters such as conversation context, agent performance metrics, and issue complexity. By modifying the parameter of recommendation frequency and timing based on real-time conditions, the system optimizes the balance between support quality and resource consumption

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If recommendations are provided with high frequency, then comprehensive guidance is offered, but clickthrough rates decrease due to agent distraction

Engineering Contradiction:

Improveguidance comprehensivenessVSAvoidrecommendation acceptance rate

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system performs preliminary prediction of recommendation usefulness before provision, analyzing conversation context and agent behavior patterns. This preliminary action ensures that only potentially valuable recommendations are provided, maintaining comprehensiveness while improving acceptance rates by avoiding distracting recommendations

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system dynamically adjusts recommendation provision based on real-time conversation state and agent response patterns. By making the recommendation system adaptive rather than static, it maintains comprehensive guidance while improving reliability of acceptance through context-aware timing and content selection

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS12278931B2Systems and methods for adaptive computer-assistance prompts

Publication Date: 2025.04.15 VERIZON PATENT & LICENSING INC

AI summary

In some implementations, an agent-assistance platform may receive a first call stream. The agent-assistance platform may identify a first topic of the first call stream. The agent-assistance platform may output, using a recommendation model, one or more recommendations associated with the first topic of the first call stream. The agent-assistance platform may determine whether a recommendation, of the one or more recommendations, is selected for inclusion in the first call stream. The agent-assistance platform may update the recommendation model with information indicating whether the recommendation is selected for inclusion in the first call stream. The agent-assistance platform may receive a second call stream. The agent-assistance platform may identify a second topic of the second call stream, the first topic being the same as the second topic. The agent-assistance platform may selectively output, using the updated recommendation model, the recommendation for the second call stream.