Summary of Key Points
Low-code process orchestration platforms like Coze, Dify, and n8n once gained popularity by helping developers avoid redundant code, enabling business users to create processes, and reducing costs for managers. However, they are now facing competition from simpler AI programming agents (Coding Agents). Nevertheless, the two are not in a zero-sum situation: low-code platforms retain their strengths of stability and controllability during transformation, while Coding Agents focus on ease of use and flexibility. In the future, a hybrid architecture will emerge, with Workflows for clear tasks and Agents for more complex issues. The ultimate competitive factor will be the ability to deeply understand a company's business knowledge (KnowHow).
I. The Highs and Lows of Low-Code Platforms: Why They Were Popular, and Why They’re in a Difficult Position Now?
Highs: 2023 was a golden year for Coze and Dify. Although AI models were advanced at the time, they were not easy for non-experts to use in product development, and developers still had to write a lot of redundant code to integrate various tools and systems. Low-code platforms solved three main problems:
- Developers: No need to write repetitive code for small tasks.
- Business Users: Could create processes effortlessly through drag-and-drop interfaces.
- Managers: Saw potential cost savings and efficiency improvements.
For example, companies used Dify for private deployments, and individuals relied on the Coze ecosystem to quickly create demos.
Lowers: These platforms were limited to simple use cases and struggled with complex business scenarios:
- Complex Processes: Highly intricate processes, such as those in customer service, required meticulous maintenance by just a few people.
- Model Limitations: AI models often provided inaccurate or irrelevant answers when used for complex tasks like managing internal documentation.
As a result, their role in production environments was limited to creating demos; they were not effective for actual work.
II. How Coding Agents Are Competing: Overcoming the Limits of Drag-and-Drop Tools
User needs have evolved—people no longer want to spend time organizing processes or using cumbersome drag-and-drop interfaces. AI programming agents (Coding Agents) have emerged:
- Early Versions: Could generate processes with a single command but were unstable and prone to errors.
- Current Versions: Include “Skill Specifications” (descriptions of processes in natural language), significantly improving stability, and have become more collaborative—AI generates the process, followed by human review and optimization.
Coding Agents have redefined the core advantages of low-code platforms: lower barriers to entry and higher efficiency. For instance, Claude Code can generate customer feedback classification processes instantly, much faster than manual drag-and-drop methods.
III. The Struggle of Low-Code Platforms to Survive: Moving from Pure Drag-and-Drop to Hybrid Models
Coze, Dify, and n8n are adapting:
- Coze3.0: Features an “online Agent team portal” that doesn’t require knowledge of local environments or code repositories, targeting users who need to quickly create demos.
- Dify: Positioned as an open-source platform for building agent-based workflows, integrating agent capabilities into low-code solutions.
- n8n: Highlights visualization and controllability with the AI Workflow Builder, allowing AI-generated processes while maintaining human approval and control over steps.
The idea is to combine the flexibility of drag-and-drop with the stability of AI to create a more versatile solution.
IV. The Future Divide: When to Use Workflows, When to Use Agents?
There’s a consensus in the industry that these tools complement each other rather than replacing one another:
- Workflows: Ideal for tasks with clear requirements and 100% stability, such as financial reimbursement processes or data synchronization.
- Agents: Suitable for open-ended tasks that require dynamic decision-making, like complex customer service or market research.
- Hybrid Architecture: Uses agents for simple issues and workflows for more complex ones, with manual intervention where needed.
The key is to allocate control appropriately—workflows handle predictable tasks, while agents handle the unpredictable.
V. The Ultimate Competition: Understanding Business Knowledge
Top AI companies (OpenAI, Anthropic) are hiring FDE (Finance and Data Engineers) who understand both technology and business. Why? Because AI models are powerful enough, but the real challenge for businesses is to organize their business knowledge effectively. For example, how to convert customer complaint processes into understandable rules for AI or turn internal documents into usable knowledge bases.
The future of tools lies in their ability to help companies capture and utilize their business expertise. Platforms like Coze or Dify that can integrate industry-specific templates (e.g., e-commerce or manufacturing processes) will have a competitive advantage, as understanding business is crucial.
Conclusion: Coze, Dify, and n8n won’t disappear; they will evolve into tools that combine AI with Workflows. Coding Agents won’t completely replace them either. The winner will be the one that can effectively help businesses apply their knowledge to practical tasks. After all, even the best tools are useless without a deep understanding of business operations.