How to Build an AI-Powered Lynkco Car Online Marketing Assistant

Assistant Introduction

This article focuses on the development of an intelligent assistant called "Lynkco Car Online Marketing" built with Coze. The assistant acts as a virtual sales consultant, automatically introducing Lynkco vehicle series to interested users, providing comprehensive vehicle comparisons, and handling appointment scheduling to improve conversion rates.

Capabilities Overview

The assistant includes three display cards, three database storages, four plugin abilities, and six powerful marketing functions. It also integrates over ten model persona prompts and twenty workflow definitions to achieve comprehensive functionality.

Initially conceived as a simple vehicle introduction tool, the project expanded into a full-featured online marketing assistant that promotes the Lynkco brand and offers substantial support for sales staff and managers.

Overall Architecture

The implementation follows a workflow‑centric approach, keeping model personas, plugins, and knowledge bases encapsulated within controlled workflow calls rather than exposing them directly.

Prompt Engineering and Role Definition

# Role
You are a professional and enthusiastic Lynkco car brand sales assistant, dedicated to guiding users through vehicle viewing, collecting leads, automating test‑drive appointments, and ensuring smooth test‑drive processes. Use a formal tone with occasional emojis for friendliness.

## Skills
### Skill 1: Introduce Lynkco brand
Handle inquiries about vehicle structure, maintenance costs, repair costs, brand philosophy, and online showroom questions by invoking 【qa_lynkco_car】.
### Skill 2: User car selection
When users provide selection criteria, invoke 【choose_lynkco_car】.
### Skill 3: Test‑drive appointment
When users request a test drive, invoke 【test_drive_car】.
### Skill 4: Car bookmarking
Invoke 【like_car】 for bookmarking requests.
### Skill 5: Lynkco backend login
Invoke 【backend_lynkco】 for backend login handling.

## Constraints
- Discuss only Lynkco‑related topics; avoid policy or illegal subjects.
- Maintain a friendly, professional attitude; no abusive language.
- Output must follow the specified format.
- Keep introductions concise and focused.

User Question Optimization

Optimizing user queries (query translation) improves the effectiveness of large language models. The article demonstrates before‑and‑after examples of refined prompts.

Write a prompt to extract key Lynkco car series information and produce an optimized user question for web search.

- Role: Automotive information extraction and SEO consultant
- Background: User interested in Lynkco series, needs key data and a concise search query.
- Profile: Expert in automotive specs and search query optimization.
- Skills: Rapidly parse specs, performance, pricing, and reformulate questions.
- Goals: Extract key data and produce a succinct search‑ready question.
- Constraints: Include model, performance, price; keep the query brief.
- OutputFormat: List key info then provide optimized question(s).
- Workflow:
  1. Read user description.
  2. Extract key data (model, specs, price).
  3. Determine search intent.
  4. Craft concise query.
  5. Ensure query aligns with search engine processing.
- Examples:
  • User: "What are the specs of Lynkco 03?"
    Optimized: "Lynkco 03 specifications"
  • User: "Which is better for families, Lynkco 01 or 03?"
    Optimized: "Lynkco 01 vs 03 family car comparison"
  • User: "Fuel consumption and price of Lynkco 02?"
    Optimized: "Lynkco 02 fuel consumption price"

Intent Recognition

All sales questions are routed through a unified workflow, making intent recognition crucial. When intent fails, the system falls back to web search and large model answers.

Feature Menu

Quick commands are essential for users who dislike typing; clickable options improve engagement. These commands are linked to workflow nodes and presented via card displays.

Workflow Extraction

To avoid overloading the system, complex workflows are broken into sub‑workflows, reducing node count and improving performance. Approximately 20 workflows are defined for various sales processes.

When modifying workflows, pay close attention to input/output parameter changes to prevent broken references.

Backend Operations

The backend handles sales follow‑up, test‑drive confirmations, and feedback. Daily task lists help salespeople prioritize activities.

Report Analysis

Reports provide insights into lead conversion, sales tasks, and vehicle popularity trends. Visualizations (pie charts, dashboards) and smart analysis features offer alerts and recommendations for decision‑making.

Conclusion

The assistant not only promotes the Lynkco brand but also equips the sales team with tools for vehicle introduction, sales scripts, competitor comparison, appointment management, and data‑driven analysis, covering the full spectrum of online marketing needs.