When to use the New Agents vs A workflow with AI

How to enable agents on self-hosted instances 👉 https://go.n8n.io/pgNQWO If you're using n8n Cloud, agents are rolling out in phases. Stay on the latest version and watch for the Agents tab on your instance over the coming week. More details here 👉 https://go.n8n.io/jmXUNm In this video, @liammcgarrigle explains where first-class agents fit within n8n's AI offering, when to use them instead of a workflow, and how the two work together. Agents are useful for tasks without fixed, predefined steps. Give an agent objectives, personality, skills, and tools, then let it decide how to complete the task. You can limit which tools it can access or require human approval before certain actions. Agents can also work where your team already does, for example in Slack. Workflows give you deterministic logic that runs the same way every time. You can also add AI to a workflow for specific steps such as classification, filling out data, or tasks that need tools and multiple layers of AI. You don't have to choose one or the other. Workflows can become tools for agents, and agents can be called from the workflow canvas. This lets you combine agent autonomy with visible, controlled workflow logic. Links: Agent docs: https://go.n8n.io/GZQNxZ How to upgrade on cloud: https://go.n8n.io/k6rqHm How to upgrade on self-hosted: https://go.n8n.io/gCvWzq Sign up for n8n cloud: https://go.n8n.io/MSCs3B Get help: https://go.n8n.io/iGppo7 Timestamps: 00:00 - Credentials, workflows, and agents 00:27 - What agents are 00:46 - Building a marketing assistant 01:16 - Deterministic workflows 01:50 - Using AI for specific workflow steps 02:21 - Agents vs agent nodes in workflows 02:48 - Combining agents and workflows 03:03 - Using agents inside workflows 03:29 - Getting access to agents