Event Focus: AI Coding Expands Beyond Code Generation

AICon Global Artificial Intelligence Development and Application Conference will be held in Shenzhen on August 21-22. Li Weining, head of internal open source at HSBC Technology, is confirmed to speak in the track “AI-native paradigm: Coding Agent reshapes the full software development process.” His session, titled “From Code Generation to an R&D Closed Loop: AI Coding Practices in the Fintech SDLC,” will focus on how AI Coding can be applied across enterprise software delivery rather than used only as an individual productivity tool.

SDLC, or software development life cycle, refers to the full process of software work, including requirements, design, coding, review, testing and delivery collaboration. In many teams, AI Coding still means code completion, snippet generation or error explanation. The fintech scenario raises a harder question: how can these capabilities operate with enough context, quality control, security and compliance to support real enterprise workflows?

Internal Open Source and Reusable Agent Skills

According to the conference information, HSBC Technology’s practice is built around internal open source and community co-creation. The idea is to collect experience from different teams across requirement analysis, architecture design, implementation, code review, testing and delivery collaboration, then turn effective practices into reusable and governable tools and Agent Skills.

An Agent Skill can be understood as a task-oriented capability package for an AI agent. It may include prompts, workflow rules, contextual requirements and ways to interact with development tools. This matters because scattered prompts and personal habits are hard to scale in a large organization. Internal open source gives teams a mechanism to share what works, improve it collectively and move from individual AI usage to organization-level AI-assisted engineering capability.

Li Weining’s background aligns with this topic. He has worked in fintech for 15 years and has experience in development, testing, operations, architecture, project delivery and product management. His past work spans HSBC Technology, GAC Automotive Finance, Xpeng financing and leasing, Hang Seng Bank and other financial or technology organizations.

Tool Integration: Bringing Agents Into the Workflow

The session will also cover integration with MCP, VS Code, GitHub Copilot, Jira and Confluence. MCP can be simply described as a mechanism for connecting models with external tools, data and development context. Its role is to help agents move beyond a chat interface and participate in existing engineering workflows.

The announced agenda covers multiple SDLC stages:

• Requirements: using Jira and Confluence to support requirement understanding and clarification;
• Design: assisting architecture proposal generation, impact analysis and technical decisions;
• Coding: working with VS Code and GitHub Copilot to improve development efficiency;
• Review: supporting code review, risk identification and standards checking;
• Testing: assisting test case generation, defect analysis and validation loops.

This framing shows a shift in AI Coding’s role. It is no longer only a code generator; it is becoming a workflow assistant that connects requirements, code, tests and knowledge bases. For fintech organizations, that connection is especially important because business rules are complex, system boundaries are sensitive, and audit and compliance requirements are strict.

Governance, Risk and Scaling

The disclosed material also highlights practical challenges. Model outputs can be unstable and may contain hallucinations, making code quality and consistency difficult to guarantee in complex business scenarios. Security and compliance risks are also present, including possible exposure of sensitive data and unauthorized tool invocation. Scaling is another challenge because teams may have different technology stacks, processes and ways of measuring value.

For that reason, the talk will discuss security governance, risk assessment and controlled adoption. Topics include data boundaries in fintech, reuse of Agent Skills, permissions, auditability and quality governance, as well as the balance between engineering efficiency and engineering risk. The agenda also mentions a path from small internal open source projects to an AI Coding platform shared and co-built by more than ten thousand people, with emphasis on selecting high-value pilots, developer adoption, training feedback and community operations.

Industry View: Engineering Becomes the Main Battleground

The AICon Shenzhen agenda is now fully online. The event includes 10 themed forums, one hands-on lab, nearly 60 sessions and more than 50 senior experts, covering Agent engineering, large model infrastructure, AI-native development, embodied intelligence, Agent security and other topics. The program reflects a broader industry shift: competition is moving from model capability alone toward reliable agents, engineering systems and business integration.

For fintech companies, the value of AI Coding will not be defined only by how much code it can produce. The more important test is whether it can work reliably, reuse knowledge, respect governance rules and leave an auditable trail inside organizational processes. The next stage is likely to treat AI Coding as part of R&D infrastructure: connected to developer tools on one side, linked to requirements, knowledge bases and workflow systems on the other, and constrained by permissions, audit and quality standards in between. Companies that can turn personal productivity gains into governed organizational capability will be better positioned for AI-native software development.