Dianxiaomi Enterprise Edition Agent use cases
Use the Dianxiaomi Enterprise Edition Agent to train and optimize brand-specific AI customer service models, and to integrate and validate custom intelligent Q&A workflows.
Scenario 1: Train and optimize brand-specific models
Background
A well-known brand has established its own AI training team and has accumulated a large conversational corpus from its private domain operations. To enhance the customer service experience on the Taotian platform, the brand wants to train an AI customer service model for its Dianxiaomi pre-sales system that reflects its unique business characteristics and communication style.
Key requirements
Develop pre-sales conversational capabilities that align with the brand's persona, tone, and service standards.
Use a high-quality conversational corpus from private domains to fine-tune two key models: the pre-sales Planner, which generates service strategies, and the pre-sales Critic, which evaluates and refines response quality.
Establish an end-to-end workflow that covers dataset upload, model training, performance evaluation, and prompt management.
Capabilities
Upload and train with your own dataset
You can securely upload your proprietary conversational corpus to use as a dedicated training dataset. The system automatically cleans and labels the data to fine-tune the pre-sales Planner and pre-sales Critic models, ensuring that the generated content aligns with your brand's context and service guidelines.
Model training and version management
Use the visual interface to configure training tasks. You can initiate separate training tasks for the pre-sales Planner or pre-sales Critic models and run parallel training to compare multiple model versions, enabling rapid iteration and optimization.
Automated evaluation and report generation
After training is complete, the system automatically performs a multi-dimensional automated evaluation, assessing factors such as relevance, compliance, brand consistency, and intent accuracy. It then generates a structured evaluation report, allowing you to visually review the model's performance in typical scenarios and identify areas for improvement.
Reference official prompts and manage custom prompts
The system includes official Dianxiaomi-recommended prompt templates for the pre-sales Planner and pre-sales Critic, which you can use as a starting point for training. You can also edit, save, version, and reuse prompts. Precise prompt engineering helps you further align the model's output with your brand's specific tone and style.
Scenario 2: Integrate and validate custom Q&A workflows
Background
A brand with mature AI development capabilities has built its own knowledge base, custom workflows, and a tailored large-scale model. The brand wants to deeply integrate its proprietary AI capabilities into specific business scenarios on the Taotian platform, such as consultations for high-value products or complex after-sales guidance, to build an end-to-end, controllable intelligent Q&A service. This integration creates a highly customized pre-sales conversational experience by flexibly orchestrating the brand's own tools and models.
Key requirements
Embed the brand's proprietary Planner and Critic models into the Dianxiaomi conversational flow.
Connect to internal knowledge bases and business tools.
Support end-to-end testing and performance monitoring of the entire conversational workflow.
Maintain full control over prompts, tool-calling logic, and decision-making paths.
Capabilities
Orchestrate custom pre-sales conversational workflows
The platform provides an open architecture that allows you to insert your own AI components into the main conversational flow:
Planner & Critic Prompt Advanced Tuning: Supports uploading, editing, and version management of custom prompts, and provides fine-grained control over the model's generation strategy and evaluation criteria;
Seamlessly connect enterprise tools: Register internal business systems as callable tools using standardized APIs.
Integrated retrieval with a private knowledge base: Automatically retrieves content from connected enterprise vector databases or structured knowledge bases, such as product documents, FAQs, or policy terms, to generate accurate and compliant answers.
End-to-end query testing sandbox
The query testing sandbox is an interactive environment that lets you enter any user query, simulate the complete dialogue flow, and view the execution in real time.
Custom trace metrics monitoring dashboard
After deployment, the system continuously collects key performance metrics.