How AI Is Transforming Technical Services and Operations at JD.com

Speaker Introduction

Xu Qichen, director of JD Platform Technical Service Department, expert member of the Open Operations Alliance and core member of the Efficient Operations Community, is responsible for operations, quality, and the open technology service platform. He previously worked at Tencent, serving most of its business lines.

1. What People Think of AI

Many people associate AI with smart hardware such as Tesla's autonomous driving, but academic research focuses on algorithms, self‑learning, branches like facial recognition and speech, and overall business impact such as cost reduction and efficiency gains.

The current AI reality in most companies is still at a 0 or 1 stage: many teams are exploring AI, investing in algorithms, but have not fully realized its potential.

2. Industry and Technical Service AI Trends

2.1 Industry Trends

Organizational structures: companies like Microsoft and JD have created independent AI research institutes.

Technology service trends: cloud providers offer GPU‑distributed services to support AI infrastructure.

Productized hardware and software: AI‑enabled speakers and devices are now common, requiring massive talent investment.

Personalized recommendation technologies remain a long‑standing focus.

2.2 JD Status

JD’s AI applications span four scenarios:

Transportation and logistics – drones, unmanned warehouses, and autonomous delivery vehicles, with route‑optimization algorithms.

Finance – customer churn prediction, loan default forecasting, and risk control.

Retail – intelligent product selection, inventory management, and smart supply chain.

Internet – user profiling, recommendation systems, and advertising algorithms.

3. Essence and Prerequisites of Technical Service AI

3.1 Essence

AI must address concrete pain points, have sufficient resources (including talent), and be built on a correct planning and technical architecture to avoid duplicated platforms.

The core is to combine scenario‑specific data, continuous learning, and optimization to solve complex problems, improve decision‑making, and create value.

3.2 Prerequisites

3.2.1 First Prerequisite – Maturity Levels

Teams are classified into five maturity levels:

Wild growth – performance bottlenecks, chaotic architecture, lacking monitoring and security.

Reconstruction – basic capability management, gradual performance fixes, data tracing.

Rapid sedimentation – reasonable, layered architecture, partial automation, tool‑driven processes.

Advanced stability – high automation, strong data management, semi‑intelligent operations.

Full intelligence – unified data, automated decision‑making, algorithmic core.

3.2.2 Second Prerequisite – Foundational Platforms

JD’s foundational platforms include a unified algorithm platform (Moon‑Landing Platform), a big‑data warehouse, and a distributed resource/service platform, providing standardized, reusable algorithms and frameworks.

4. JD Technical Service Team’s Scenario‑Based Practices

4.1 User Feedback

The team built a platform to ingest millions of daily user feedback entries, applying word‑segmentation and classification algorithms to reduce manual effort by 60 % and further improve accuracy with AI models such as KNN.

After two months, the system achieved a 77 % recognition rate, later surpassing 90 % with continued model refinement.

4.2 Prediction

Predictive models use recent data (e.g., 15‑day windows) with weighted moving averages, seasonal promotion adjustments, and threshold‑based alerts, achieving low false‑positive alarm rates and supporting capacity planning.

4.3 Woodpecker System AI‑ization

4.3.1 User Feedback Scenario

AI helps detect and alert on problematic activity pages, saving 20‑40 hours per day by automatically identifying missing IDs, broken links, or mis‑configured assets.

4.3.2 Image Recognition Scenario

Image‑recognition pipelines extract visual features from activity page screenshots, using cosine similarity and clustering to verify content consistency.

5. Summary and Outlook

China’s AI development plan aims for world‑leading technology by 2025 and global leadership by 2030. The speaker emphasizes that AI will become a core driver for operations, testing, DevOps, and autonomous delivery, urging teams to align resources, architecture, and data capabilities to seize the AI wave.