Workflow or AI Agent? The 5-Question Test I Use Before Adding an LLM Node

说明与用途

五个问题的测试帮助决定在 n8n 中使用普通工作流、单个 LLM 步骤还是完整 AI 代理。结果是一个固定流程，包含一个结构化 LLM 调用、解析节点和确定性 IF/Switch 逻辑。

问题

Deciding whether to use a plain workflow, a single LLM step, or an AI agent in n8n automations.

结果

Using the five‑question test consistently produces deterministic workflows with a single LLM step, avoiding the risks of full AI agents while still leveraging LLMs for parsing messy text.

工作流

• 1. Trigger (Webhook or Schedule)
• 2. Config (Code node) – set baseUrl and model
• 3. Build Prompt (Code node) – strict system prompt with JSON schema
• 4. LLM (HTTP Request) – POST to chat/completions with response_format json_object and low temperature
• 5. Parse (Code node) – extract JSON, validate, handle errors
• 6. Decide (IF/Switch) – deterministic rules on parsed fields

证据

Every one of my workflows ends the LLM step with the same small Parse node: …

来源

If I can describe the decision as a rule a new hire could follow without judgment, it doesn't need a model.

来源

Models are good at turning unstructured text into structured fields: a review into sentiment plus topics, a meeting transcript into decisions and owners, a PDF invoice into vendor, totals and line items.

来源