Use Cases

Steps

1. 1.Configure Dify workflow and connect Brave Search to fetch industry trends.
2. 2.Set AI Agent prompts to define brand tone and content format.
3. 3.Have the Agent generate a 7-day content calendar.

Steps

1. 1.Enter target keywords and competitor domain names
2. 2.Agent automatically analyzes SERP results and user intent
3. 3.Generate SEO-optimized article outlines

Steps

1. 1.Install GitHub MCP Server in Cursor
2. 2.Set up code review prompt template
3. 3.Trigger automatic review and view suggestions

Steps

1. 1.Configure the database MCP Server to connect to the target database
2. 2.Set read-only permissions to ensure data security
3. 3.Write system prompts describing the database table structure

Steps

1. 1.Upload PDFs to a designated folder
2. 2.Configure filesystem MCP to read files
3. 3.Let AI extract key information and structure it

Steps

1. 1.List target competitors
2. 2.Agent automatically visits official websites, App Store reviews, and social media
3. 3.Analyze product feature matrix and differentiation advantages

Steps

1. 1.Create a Gmail Trigger workflow in n8n
2. 2.Add an AI node to determine email type and priority
3. 3.Generate reply drafts and mark important emails

Steps

1. 1.Upload the recording file after the meeting
2. 2.Whisper API automatically transcribes to text
3. 3.AI extracts Action Items and assignees

Steps

1. 1.Input the target category and target platform
2. 2.Agent searches for hot-selling products and market data
3. 3.Analyze competitive landscape and price range

Steps

1. 1.Enter the target company name and research focus areas
2. 2.Agent automatically searches the company website, Crunchbase, and LinkedIn
3. 3.Analyze funding rounds, valuation trends, and investors

Steps

1. 1.Configure the filesystem MCP to allow the AI to read the codebase
2. 2.Let the AI analyze the Controller/Route layer code
3. 3.Automatically identify interface parameters and return value structures

Steps

1. 1.Save resume PDFs to a designated folder
2. 2.Let AI extract education, experience, and skill information
3. 3.Automatically score and rank according to job description

Steps

1. 1.Use n8n to periodically pull new leads from CRM (HubSpot/Salesforce)
2. 2.Design scoring rules: company size × industry fit × interaction frequency
3. 3.AI generates personalized emails for each high-scoring lead (citing the client's latest company news)

Steps

1. 1.Create a 'JD Generation' workflow in Dify, input key job information
2. 2.AI generates a complete JD based on templates and industry standards
3. 3.Use browser-use to control the browser and automatically log into each recruitment platform

Steps

1. 1.Connect data sources: App Store API / CSV import / Zendesk tickets
2. 2.AI tags each feedback: feature request / bug / performance / UI experience
3. 3.Automatically cluster similar feedback and count mention frequency

Steps

1. 1.Collect student information: target exam, current level, daily available study time
2. 2.AI breaks down the exam syllabus and generates a knowledge point map with priorities
3. 3.Push daily learning tasks and corresponding resources

Steps

1. 1.Map the entire marketing chain and identify high-repetition tasks
2. 2.Select appropriate AI tools for each task
3. 3.Build a standardized prompt template library

Steps

1. 1.Install an AI coding plugin in your IDE (Cursor or Copilot)
2. 2.Provide sufficient context to the AI (project description, tech stack)
3. 3.Describe requirements in natural language and let AI generate the code skeleton

Steps

1. 1.Use Perplexity/Claude to search for competitor ranking content for target keywords
2. 2.Analyze the structure, word count, and content gaps of the top 5 articles
3. 3.Design an SEO-optimized article outline (H1/H2/H3 structure)

Steps

1. 1.Connect user feedback data sources (App Store, Zendesk, NPS, etc.)
2. 2.Design AI classification prompt, define classification dimensions and tag system
3. 3.Use Claude API to batch process feedback and generate structured data

Steps

1. 1.Map out your current sales process and identify the most time-consuming repetitive tasks.
2. 2.Integrate CRM data and configure a lead enrichment tool (e.g., Clay).
3. 3.Design personalized email prompt templates that dynamically generate based on prospect profiles.

Steps

1. 1.Organize enterprise documents: Export PDF/Markdown files from Confluence/Feishu
2. 2.Install dependencies: pip install chromadb langchain anthropic
3. 3.Write document loading script: Read files, split into chunks of 500-1000 characters, generate vectors, and store in Chroma

Steps

1. 1.Create .mcp.json in the project root, configuring GitHub MCP and filesystem MCP
2. 2.Create CLAUDE.md to define code review rules and output format (🔴🟡🟢 levels)
3. 3.Test the workflow in Claude Code with "review PR #1"

Steps

1. 1.Use RSSHub to aggregate RSS feeds from target information sources
2. 2.Set up an n8n scheduled task to trigger scraping at 7 AM every morning
3. 3.AI evaluates the importance of each piece of content (1-5 scale) and filters out low-scoring items

Steps

1. 1.Organize the product parameter table, including fields such as model, specifications, materials, and applicable scenarios.
2. 2.Design platform-specific prompt templates (Taobao focuses on emotion, JD focuses on parameters).
3. 3.Use Python to batch read parameters and call the OpenAI API line by line to generate copy.

Steps

1. 1.Data Integration: Connect ERP (Kingdee/SAP), bank statements, and e-commerce platform data to establish a unified data pipeline, with Python ETL scripts pulling data daily automatically.
2. 2.AI Analysis Layer: Send standardized data to GPT-4o to generate key metric analysis, anomaly detection, and narrative interpretation comparing with previous periods and budgets.
3. 3.Report Template: Build standard report templates in Power BI, with AI analysis results automatically populating predefined formats (executive summary + detailed schedules + trend charts).

Steps

1. 1.Current state analysis: Analyze ticket data from the past 3 months, identify the top 20 issue types, and calculate resolution time and repeat rate for each type.
2. 2.Knowledge base organization: Organize FAQs and customer service documents into a structured format (problem description - solution - relevant screenshots - applicable versions) and migrate to Notion.
3. 3.AI configuration: Enable Fin AI in Intercom, connect the Notion knowledge base, configure response style, and set trigger conditions for human handoff.

Steps

1. 1.Assess current workflow: Map time allocation across case processing stages, identify the most time-consuming repetitive tasks (contract review, legal research, client communication).
2. 2.Deploy Harvey AI: Integrate into contract review workflow; AI generates preliminary review reports, allowing lawyers to focus on strategic decisions rather than line-by-line review.
3. 3.Build a legal research database: Compile research findings on common legal issues into a knowledge base, paired with ChatGPT for quick citation retrieval.

Steps

1. 1.AI-generated sand table scripts: Generate tailored presentation scripts for different unit types, floor levels, and customer segments (first-time buyers, upgraders, investors) for sales staff to use on demand.
2. 2.AI-rendered interior images: Use Midjourney to quickly generate renderings of different decoration styles to help customers envision their future home.
3. 3.HeyGen digital presenter: Create AI-powered introduction videos available 24/7, allowing remote customers to learn about the project anytime.

Steps

1. 1.Accelerate market research with AI: Use AI to analyze competitors, generate interview guides, conduct 20 user interviews, and extract key insights with AI.
2. 2.AI-assisted product planning: Input user insights into Claude to generate a PRD (Product Requirements Document) and feature priority matrix.
3. 3.Rapid prototyping with Figma AI: Use Figma AI to generate UI framework, refine design, and complete a 20-page prototype in 1 day.

Steps

1. 1.Tool Selection Testing: Compare Midjourney, Firefly, and SD in terms of accuracy, efficiency, and cost for product image generation, and select the most suitable tool combination.
2. 2.Build Prompt Library: Create standardized generation prompts for different product categories (clothing/digital/home) to ensure consistent brand style.
3. 3.Workflow Standardization: Front-end product image → Photoshop cutout → AI background replacement → batch scene generation, forming a standard SOP.

Steps

1. 1.Build a diagnostic assessment system: Design a standardized academic diagnostic question bank (200 questions per subject). After the entrance assessment, AI automatically outputs a knowledge point mastery report and a list of weak areas.
2. 2.Generate personalized learning paths: Based on diagnostic results, AI creates targeted practice plans for each student, prioritizing weak knowledge points in high-frequency exam topics.
3. 3.AI-assisted lesson preparation: Teachers input teaching objectives into iFlytek Spark, and AI automatically generates draft courseware, practice questions, and homework. Teachers review and revise, reducing preparation time from 3 hours to 1.5 hours.

Steps

1. 1.AI demand forecasting: Integrate historical sales data (daily sales per dish), weather, holidays, and nearby events to build an LSTM prediction model. Push daily inventory recommendations (precise quantities per dish) to each store by 8 AM.
2. 2.Smart ordering system: On the mini-program ordering page, dynamically adjust dish recommendation order and display based on time of day, user order history, and current inventory to quickly move soon-to-expire ingredients.
3. 3.Personalized marketing: Analyze member purchase behavior to segment users into 5 types (frequent regulars, price-sensitive, novelty seekers, family-type, single-type). Target push coupons accordingly (new dishes for regulars, discounts for price-sensitive).

Steps

1. 1.Build trend monitoring system: Use Python to scrape popular content tags from Xiaohongshu, Douyin, and Instagram, integrate Google Trends API, and automatically generate daily 'Trending Keyword Rankings' and 'Emerging Style Direction' reports.
2. 2.AI-powered product selection analysis: Input trend reports, competitor new product data, and historical similar product sales data into ChatGPT to generate comprehensive product selection recommendations including 'recommendation rationale', 'predicted sales range', and 'risk warnings'.
3. 3.Small-batch test orders: Place initial orders of 50 units for AI-recommended products, and decide whether to restock based on 7-day sales velocity (daily average sales + repurchase rate + add-to-cart rate).

Steps

1. 1.Omnichannel Integration: Use Zendesk API to integrate Taobao Wangwang, WeChat Work, and Amazon Seller Central messages, funneling them into a single customer service workstation, eliminating the need for agents to switch platforms.
2. 2.Unified Knowledge Base: Organize product descriptions, return/exchange policies, and FAQs into a structured knowledge base (JSON format), connect to ChatGPT API, and support synchronized updates across platforms.
3. 3.AI Tiered Processing: AI automatically resolves standard issues (shipping tracking, refund status, product inquiries); complex issues (disputes, special returns, complaints) are automatically escalated to human agents with AI-generated issue summaries.

Steps

1. 1.Systematic topic selection: Every Monday, use ChatGPT to analyze trending topics across platforms, generate 20-30 topic ideas based on account positioning, filter out the 10 with the highest viral potential, and create a Notion content calendar.
2. 2.One draft, multiple uses: First write a 2,000-word in-depth article, then use AI to automatically rewrite it into: a Douyin/Kuaishou script (500 words), a Xiaohongshu post with images (300 words + 5 images), a WeChat public account article (1,500 words), and Twitter/LinkedIn posts (280 words).
3. 3.Batch image generation: Use Midjourney to generate cover images that fit each platform's dimensions, then use Canva AI to quickly adjust layouts and brand elements. Production time for each set of visual assets drops from 1 hour to 15 minutes.

Steps

1. 1.AI Customer Profile Analysis: Integrate customer source channels, browsing history, query keywords, and property viewing behavior data to train an intent prediction model, classifying customers into 'high/medium/low intent' categories and automatically assigning follow-up priorities.
2. 2.Pre-Show AI Preparation: The agent inputs property information and customer profiles, and the AI generates personalized showing scripts (tailored to different concerns: school district, investment, self-use) along with 30 common questions and reference answers that may arise.
3. 3.Real-Time Assistance During Showings: Develop a WeChat mini-program that allows agents to access AI suggestions on the fly (competitor comparison data, historical transaction prices in the same community, mortgage calculator) to answer customer questions more promptly and professionally.

Steps

1. 1.AI Resume Pre-screening: Integrate an AI scoring module into the ATS system to automatically score resumes based on key JD requirements (skill match, experience relevance, education). Only resumes scoring above 60 enter manual review, reducing pre-screening time by 70%.
2. 2.Job Description Optimization: Use AI to analyze successful hires from the past 3 years, identify candidate traits that truly predict success, rewrite JDs more precisely (reducing irrelevant applications by 40%), and check for gender/age bias in language.
3. 3.Intelligent Interview Question Bank: AI generates structured interview questions (behavioral + situational) tailored to different roles, with scoring criteria, reducing interviewer bias, and building a reference for evaluating candidate response quality.

Steps

1. 1.After switching models, manually adjust the thinking level to the level actually required by the task to avoid wasting tokens by keeping a high reasoning level.
2. 2.Periodically use /graphify or /compact to compress sessions during large projects, preventing excessively long histories from increasing the cost of each new message.
3. 3.Before complex agentic tasks, first use a cheaper model to decompose the task and clarify the scope, then hand it over to Fable for execution, reducing the number of exploratory rounds.