LLM Agent Chatbot for E-commerce Support Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional online support systems for e-commerce platforms often require users to navigate through multiple interfaces and input points to resolve inquiries, frequently necessitating contact with customer care, and lack intuitive natural language interfaces for help articles.
Innovation Solution
A system that configures a set of tools on an interface system, utilizing a large language model (LLM) to process user queries via a chatbot application. The system generates prompts that include user queries, contextual information, and available tools, extracts actions and inputs from the LLM's responses, and executes corresponding tools to provide responses to users.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional support mechanisms are used, then users can access help through multiple interfaces, but users must navigate through multiple buttons and input points to reach resolution
Solution Approach 1:
The patent introduces an LLM-based agent as an intermediary between users and the support system. This agent receives natural language queries, processes them through multiple internal tools and knowledge bases, and provides resolutions without requiring users to navigate multiple interfaces. The agent acts as a mediator that consolidates complex navigation into a single natural language interaction point.
Solution Approach 2:
The LLM-based agent serves multiple functions within a single interface: it processes natural language queries, searches knowledge bases, executes tools, and provides resolutions. This multi-functional approach eliminates the need for separate interfaces for different support functions, thereby reducing the number of buttons and input points users must navigate through.
2Loss of information
If help articles are provided, then users can find information to resolve issues, but the articles are cumbersome to navigate and lack intuitive interface
Solution Approach 1:
The LLM-based agent serves as an intermediary that provides an intuitive natural language interface to access help articles and knowledge bases. Instead of requiring users to navigate cumbersome article structures, the agent processes natural language queries and retrieves relevant information through its internal knowledge base, thereby maintaining information accessibility while dramatically improving ease of operation.
3Extent of automation
If self-service support is implemented, then users prefer autonomous assistance, but users are still routed to live agent care for assistance
Solution Approach 1:
The patent implements self-service support through an autonomous LLM-based agent that can independently process user queries, search knowledge bases, execute tools, and provide resolutions without human intervention. The agent operates continuously and can handle complex multi-step problems autonomously, thereby enhancing both self-service capability and resolution accuracy compared to traditional automated systems.
Solution Approach 2:
The system incorporates feedback mechanisms where the LLM agent iteratively processes user responses and adjusts its approach. When the agent encounters problems it cannot resolve, it can seek guidance from human agents through a structured feedback loop, ensuring high resolution accuracy while maintaining autonomous self-service capability for the majority of queries.
Data Source
AI summary
An online system provides a support application including a chatbot application. One or more tools may each be configured to access external data. The interface system hosts an agent powered by an underlying large language model. The online system receives a user query via the chatbot application. For at least one or more iterations, the online system performs steps to provide a prompt to the LLM that specifies at least the user query, contextual information, a list of available tools, or a request to output an action. The system parses the response from the LLM to extract a selected action and action inputs for the selected action. The system triggers execution of a respective tool that corresponds to the selected action with the action inputs. The system generates a response to the user query and transmits the response to the client device.