What is AI Workflow?
A predetermined sequence of steps with AI embedded at specific points — as opposed to an autonomous agent.
Definition
An AI workflow is an automation where the path from input to output is fixed in advance, with AI (typically an LLM call) embedded at one or more specific steps to handle tasks like classification, extraction, or generation. The overall flow is deterministic and human-designed; the AI handles bounded subtasks within that flow.
Example
A document processing workflow: trigger on new email → LLM step: classify document type → route: if invoice go to step 4, if contract go to step 6 → extract fields → write to spreadsheet. The LLM classifies; a human designer decided every possible path.
AI Workflow vs ai-agent: What's the difference?
A predetermined sequence of steps with AI embedded at specific points — as opposed to an autonomous agent.
A workflow follows a fixed path a human defined. An agent decides its own path based on what it finds. Workflows are more predictable; agents handle more complexity.