AI-Driven Supply Chain Optimization at JD.com

Key Concepts

• AI-driven Supply Chain Optimization: Using artificial intelligence to improve efficiency and decision-making in supply chain processes.
• Chatbot Interface: A conversational interface for users to interact with the supply chain system using natural language.
• Explainable AI (XAI): AI models that provide transparency and justification for their predictions and decisions.
• Synthetic Data: Data generated artificially to augment real-world data for training AI models.
• Interactive Optimization: Real-time fine-tuning of data inputs and solutions in optimization problems through user interaction.
• Digital Twin: A virtual representation of a physical system (e.g., a supply chain) used for simulation and analysis.
• Demand Forecasting: Predicting future demand for products to optimize inventory and production planning.
• Global Optimization: Optimizing processes across the entire supply chain, rather than in isolated segments.

JD.com Overview

• JD.com is China's largest retailer by revenue and a leading technology and service provider with supply chain at its core.
• Serves approximately 600 million users and offers over 10 million products in its first-party business.
• Operates more than 1,500 self-operated warehouses.
• Fulfills over 90% of orders on the same day or the next day.
• In 2024, net revenue exceeded $115 billion USD.
• Faces challenges such as fluctuating demand, especially during peak shopping events like the 618 shopping festival.

JD.com's Business Offerings

• JD Retail: Connects brands and customers, serving over 600 million users.
• JD Logistics: Provides delivery services, ensuring safe and efficient goods delivery.
• JD Technology: Offers foundational services and solutions like cloud and financial services.
• JD Property & JD House: Other business segments contributing to the JD ecosystem.

Conventional vs. AI-Based Supply Chain

• Conventional Supply Chain:
  • Involves key stages: marketing, planning, inventory, and fulfillment.
  • Challenges: fragmented systems, manual processes, lack of trusted data, inefficiencies, miscommunication, and slow response times.
  • Often divided into multiple segments handled by separate departments with limited data flow.
• AI-Based Supply Chain:
  • Interactive, integrated, and collaborative system leveraging AI for planning, replenishment, and order fulfillment.
  • Dynamically combines capabilities based on user needs, ensuring global optimization.
  • User interactions occur through chatbots, enabling intuitive and efficient communication.

AI Applications at JD.com

Interactive Chatbot

• Users can ask questions about inventory levels, future sales, etc., using natural language.
• The AI chatbot understands the question, identifies the required information, and calls different "agents" (forecasting, planning, optimization) to gather data and provide answers.
• Enables real-time communication and decision-making.

Demand Forecasting

• Utilizes private data from 600 million active users, public data (weather, market trends, tariffs), and synthetic data.
• Synthetic Data Generation: Analyzes past transactions to understand successful actions and generates similar data for future forecasting.
• Self-learning capability: AI tools learn from historical data, leading to a 15% increase in prediction accuracy.
• Technical paper available detailing the use of patch neuronet networks to find patterns.

Explainable AI (XAI)

• Addresses the challenge of convincing stakeholders to trust AI-driven forecasts.
• Explains the major factors influencing the forecast to stakeholders.
• Identifies key factors like promotion, seasonality, and market trends.
• Provides a range of values instead of a single number for more accurate predictions.
• Example: During sales events like 618, XAI helps companies understand how to reach their sales targets by analyzing historical performance, demand fluctuations, and promotion strategies.
• Presents different promotion plans with varying costs and outcomes.

Interactive Optimization

• Addresses the complexity of traditional optimization processes (coding, data cleaning, parameter guessing).
• Users can describe the optimization problem in natural language (e.g., "I want to order 1,000 units and put them into my warehouses using the cheapest method").
• The AI converts the English input into a mathematical model, automatically generates code, grabs relevant data, and solves the problem.
• Supports "what-if" analysis by allowing users to quickly adjust parameters and re-solve the problem.

Key Factors for Successful AI Implementation

• Team: Building a strong team of talented engineers and data scientists.
• Leadership: Having a leadership team that embraces AI and facilitates its integration into the organization.
• Data: Leveraging rich and diverse data sources to train AI models.
• Organizational Change: Creating a flat organizational structure that allows AI engineers to work directly with business units.
• Collaboration with Academia: Partnering with universities for research and talent acquisition.

Awards and Recognition

• Informs Edelman Award Finalist.
• Informs Prize (highest award for applying OR and data analytics).
• Wagner Prize for technical problem-solving.

AI Tsunami and Trade-offs

• AI is rapidly evolving, requiring companies to make trade-offs between model size, cost, and efficiency.
• JD.com focuses on using smaller models that leverage experience and data to solve specific problems.
• Achieves a balance between effectiveness, efficiency, and cost.

Live Streaming

• Live streaming is a significant e-commerce trend in China.
• JD.com acknowledges its impact but emphasizes the importance of a robust supply chain, which live streaming platforms often lack.
• JD.com's CEO has also participated in live streaming events.

Organizational Impact of AI

• AI is expected to have a significant impact on organizational structures, automating functions and enabling faster feedback.
• JD.com aims to share its AI capabilities with smaller companies through APIs.
• Effective leadership is crucial for embracing AI and driving organizational change.

Talent Acquisition

• JD.com actively recruits talent from universities, including a "genius" program.
• Focuses on hiring individuals who can communicate effectively across different teams and drive cultural change.
• Prioritizes well-rounded individuals who can understand and contribute to multiple aspects of the business.

Explainable AI (Detailed)

• Two approaches:
  • Interactive explanation: Facilitates real-time discussions and adjustments based on input from different units.
  • Post-results explanation: Provides weights and factors influencing the results, allowing users to understand the reasoning behind the recommendations.
• Aims to reconcile automated recommendations with domain expert knowledge through collaboration and data-driven decision-making.

Synthetic Data Generation (Detailed)

• Addresses the issue of dirty and missing data in raw datasets.
• Analyzes historical ordering data to identify successful decisions and patterns.
• Generates synthetic data based on these patterns, introducing variations to create a more robust dataset.
• Filters out bad data from the original dataset to improve the overall quality of the training data.

Handling External Shocks (e.g., Tariffs)

• JD.com uses its data intelligence and inventory repositioning capabilities to buffer against disruptions.
• Maintains a high inventory level to ensure products can be shipped from the closest warehouse, minimizing delivery times.
• Simulates future scenarios using a digital twin to stress-test the supply chain and identify potential vulnerabilities.

Predicting Customer Behavior

• JD.com can forecast customer behavior at an individual level using membership data, purchasing history, and similar purchasing patterns within their group.
• Uses a "10,000 people, 10,000 faces" campaign to create a digital twin of each customer, capturing their unique behavior.

Top 3 Lessons for Other Large Companies

1. Talent: Hire the right people with the necessary skills and expertise.
2. Leadership: Embrace AI and facilitate its integration into the organization.
3. Coordination: Foster collaboration among different units to ensure a unified approach to AI implementation.

Synthesis/Conclusion

JD.com is leveraging AI to transform its supply chain operations, focusing on interactive communication, explainable AI, and data-driven decision-making. The company emphasizes the importance of building a strong team, fostering a culture of innovation, and leveraging its vast data resources to achieve a competitive advantage. By balancing effectiveness, efficiency, and cost, JD.com is positioning itself as a leader in the AI-driven e-commerce landscape.