AI Workflow
Also called LLM workflow, Predefined workflow
A sequence of steps, some run by AI models, that follows a path set in advance by the builder. Unlike an agent, it does not decide its own steps.
What is an AI workflow?
An AI workflow is a fixed series of steps in which AI models and other tools are orchestrated through predefined code paths. The builder decides the order of the steps in advance, and the model fills in the parts that need language or judgement, such as summarising a document or classifying a message.
How is it different from an agent?
In Anthropic's framing, workflows follow paths set by code, while agents let the model direct its own process and tool use. Workflows are more predictable and easier to test. Agents are more flexible but harder to control. An agentic workflow sits closer to the agent end, with the model choosing steps as it goes.
What is a simple example?
A support tool receives an email, an AI model classifies it, a second model drafts a reply for that category, and a person approves it before it is sent. The steps always run in that order.
Why does it matter?
For many business tasks a fixed workflow is cheaper, faster and safer than a free-roaming agent. Common patterns include prompt chaining and routing, and people are often kept in the loop with human review.