Agents versus fixed workflows
n8n’s documentation describes an agent as "a chain that knows how to make decisions" (n8n Docs). A normal workflow runs the same steps every time. An agent uses a language model to choose its next step, and it can call tools: add-ons that let it fetch information or act in another system.
A small business example
A customer messages a shop on WhatsApp: "Is the black kurta in medium still available, and can it reach Faisalabad by Friday?" An agent could:
- Read the message and work out it needs stock and delivery details.
- Check the stock sheet for the item and size.
- Check the courier’s delivery times for Faisalabad.
- Draft a reply and pass it to a person to approve before it is sent.
Keep a person in the loop
Agents can misread a request or act on a wrong answer, just as the underlying model can. For anything that spends money, changes records or talks to customers, have the agent draft and a person approve. That keeps the time saving without the risk.