AI automation puts an AI model to work inside an automated process. Typical examples are classifying incoming emails, extracting fields from invoices, drafting replies for review, or summarizing every new support ticket and posting it to a channel.
Most useful AI automation is a normal workflow with AI doing the one step that used to need human judgment, while ordinary logic handles the rest. That keeps it predictable. Tools such as n8n are often used to wire these steps together through webhook triggers and API calls.
Because model output can be wrong, well-built AI automation checks the result, handles failures with error handling and retry logic, and sends uncertain cases to a person.
Related: Workflow Automation, AI Agent, Agentic AI.