Generative AI in Software Engineering: Trends, Low‑Code Integration, and Team Workflow

In 2024, generative AI (Generative AI) has become increasingly pervasive in software engineering, enhancing traditional development and delivery processes. AI‑code capabilities are infiltrating every stage of the workflow, including modeling, coding, testing, and documentation, while low‑code platforms are exploring how to combine with large‑model abilities to meet rapid customer‑driven customization demands.

The impact of AI‑code on software development can be summarized in three aspects: (1) evolving from single‑point coding assistance to supporting the entire software lifecycle; (2) merging AI‑code with vertical low‑code/no‑code scenarios to improve content production and functional delivery; (3) spreading from individual developers to team workflows, gradually becoming a core component of enterprise IT infrastructure.

Traditional low‑code faces challenges. The NetEase Cloud Music team introduced Tango , a low‑code solution that does not replace source‑code development but builds a visual interface on top of the codebase, reducing the need to focus on programming languages and frameworks while providing Copilot‑like assistance and allowing custom extensions.

Professional low‑code platforms offer highly encapsulated visual interfaces, which can hinder rapid customization, lack advanced server‑side abstractions, and limit flexible extensions. Complex cross‑service flows become cumbersome in visual editors. Leveraging large‑model capabilities also incurs extra costs due to conversion pipelines; training private small models is expensive and often less effective than using large, parameter‑rich models.

Code remains the core technical asset. AI‑code generation still relies on massive code corpora for training (e.g., GitHub Copilot’s Codex). Continuous evolution of AI‑code requires human developers and community‑driven mainstream technologies to maximize its utility.

Several AI‑driven full‑lifecycle management tools are highlighted: GitHub Copilot (with extensions), Replit (cloud collaborative IDE), Google Project IDX (AI‑assisted online IDE), and Sonar (AI‑enhanced code quality and security). Each integrates AI into different workflow nodes to boost developer experience and efficiency.

AI also powers vertical scenario solutions: Quest AI (design‑to‑code), HTTPie AI (API testing assistance), Vercel V0 (AI‑generated front‑end UI), among others, demonstrating how domain‑specific models can streamline specialized tasks.

Team‑oriented AI tools include GenPen.AI (project collaboration and code generation), CodeMaker (NetEase’s internal intelligent R&D workstation offering code completion, chat, generation, search, review, and scanning), and LangBase (AI application management and orchestration platform). These solutions aim to embed AI deeper into collaborative development processes.

Within NetEase Cloud Music, a progressive low‑code strategy has been pursued since 2023: starting with GitHub Copilot and Tango to enhance individual workflows, then in 2024 expanding AI‑code into team workflows across front‑end and back‑end scenarios. Integration with LangBase agents (DevAgent, DesignAgent) and custom code‑completion models aims to replace Copilot with tailored solutions, focusing on four pillars: custom completion models, CM plugin extensions, workflow extensions, and AI‑driven wizards.

The conclusion emphasizes that AI is evolving faster than anticipated. While fully automating software engineering remains premature, AI‑code is moving from personal assistance to team‑wide integration, gradually becoming a pivotal assistant in every core development node, thereby improving development quality and efficiency.

References: [1] https://github.com/wwsun, [2] https://github.com/NetEase/tango, [3] https://www.figma.com/community/plugin/1174548852019950797/seal-ai-powered-figma-to-code-react-rn-vue-html-d2c