AI Agent-Driven Open Unmanned Retail Network

📌 Background

After going through technological iterations such as RFID and gravity sensing, the unmanned retail industry is entering the "Agent Era" in 2026. Driven by large language models for dynamic product selection, replenishment, and pricing algorithms, combined with computer vision and embodied AI robots, this approach solves the pain points of traditional unmanned cabinets—such as high loss rates, high operating costs, and limited SKUs. Companies like Feng E Zu Shi have achieved large-scale commercialization, marking the shift of this model from scenario experimentation to a scaled profitability stage, making it a core growth driver for offline retail digitalization.

👤 Target Customers

Enterprise clients (such as property management or administrative departments in semi-enclosed scenarios like offices, factories, and schools); consumers pay indirectly.

💰 Revenue Streams

1) Product Sales Margin: Increase inventory turnover through intelligent product selection to earn gross profit on sold goods; 2) Data and Advertising Service Fees: Leverage cabinet screens and user consumption data to charge brand owners for targeted marketing and product recommendation fees; 3) Franchise and Equipment Leasing: Collect equipment deposits, technical service fees, or SaaS subscription fees from partners.

🧮 Cost Structure

Hardware costs (AI vision cabinets, sensors, robots), R&D costs (AI model training and inference), supply chain costs (product procurement and cold-chain delivery), replenishment and operations labor costs, location rent or revenue sharing.

🛡️ Moat

1. Scale Network Effect: A massive network of locations, such as 180,000 shelves, builds extremely high operational experience and data barriers. 2. AI Algorithm Advantage: Product selection and pricing algorithms trained on massive consumption data continuously optimize gross profit and turnover rate. 3. Supply Chain Integration Capability: Deep cooperation with fast-moving consumer goods (FMCG) brands to achieve low-cost, direct sourcing from the origin.

🔑 Keys to Success

• Quickly capture a large number of high-quality locations and achieve scale effects
• Continuously invest in algorithm R&D to maintain leadership in product selection strategies and dynamic pricing
• Build a win-win data and advertising monetization model with brand owners

⚠️ Risks

• Excessively high location land-grabbing costs making it difficult for the profit model to run smoothly
• Large model decision-making errors causing large-scale merchandise loss or safety incidents
• B-end customer concerns about data privacy limiting commercialization potential

🏢 Cases

• Feng E Zu Shi: Covering over 50 cities nationwide and operating more than 180,000 intelligent vending shelves, its equipment utilizes AI large models to achieve dynamic product selection and smart replenishment, having processed over 600 million orders cumulatively and currently sprinting toward an IPO on the Hong Kong Stock Exchange.
• Youbao Online: A well-known domestic intelligent retail service provider that also deploys smart cabinets on a large scale in government and enterprise offices, factories, and other scenarios, optimizing operational efficiency through AI decision-making.

📊 SWOT Analysis