How AGI and AIGC Are Transforming Software Development at NetEase Cloud

Introduction

At the 2023 EE (Excellence Engineering Productivity) Summit, NetEase Cloud's CTO Xu Hangsheng delivered a talk titled "AGI in NetEase Cloud: Technical Efficiency and Business Innovation," which this article documents.

AGI Overview

AGI (Artificial General Intelligence) is the ultimate goal of AI research. While true AGI remains distant, current AIGC (AI‑generated content) technologies such as large language models are already reshaping product development.

AIGC in the Software Development Lifecycle

NetEase Cloud provides a PaaS platform with SDKs for IM and RTC services. AIGC can be applied to every SDLC stage:

Requirement analysis: Use text generation to create market research, user stories, and documentation.

Product design: Generate interaction designs, system architecture diagrams, and flowcharts.

Development: Leverage code generation, refactoring, explanation, test‑case creation, and synthetic data production.

Testing, deployment, and operation: Apply AI for performance analysis, intelligent documentation, and automated customer service.

Practical Tips for Using AI Models

Provide clear, unambiguous instructions. Ambiguity leads to incorrect token‑level operations.

Allow the model time to think. Use Chain‑of‑Thought prompting to break tasks into steps.

Supply examples. One‑shot or few‑shot prompts improve model understanding.

Enable the model to ask clarifying questions. This ensures accurate outputs when information is insufficient.

Copilot‑Style Applications

Inspired by Microsoft’s Copilot, NetEase integrates AI assistants into code review (GitLab MR AI Review), code completion, and rapid prototyping, positioning AI as a supportive co‑pilot rather than a replacement.

Knowledge Bases and Customer Support

Three knowledge‑base layers—product capabilities, customer information, and personal role knowledge—are enriched with AI to improve document search, automate issue triage, and create a 24‑hour AI assistant for developers.

Case Study: Annotation Assistant

For video conferencing background‑blur challenges, NetEase combines BLIP, Stable Diffusion, Grounding‑DINO, and SAM to generate synthetic training data and automate labeling, cutting annotation time and cost while maintaining high accuracy.

Conclusion

By embedding AIGC throughout the SDLC, NetEase Cloud reduces repetitive coding, accelerates design, and enhances developer experience, demonstrating how AI‑driven productivity tools can reshape software engineering and deliver measurable efficiency gains.