Workflows

A workflow is how you make a repeatable operation run on its own. Where the Agent brings judgment to work that varies, a workflow brings reliability to work that shouldn't: the same steps, in the same order, every time.

Triggers and actions

Every workflow is built from two kinds of building block:

• A trigger is what starts it, such as a schedule, a file arriving, or a change in your data.
• Actions are the steps it runs once triggered, such as reading a file, transforming records, or creating a task.

Chaining a trigger to a series of actions is all a workflow really is. The power comes from what those actions can touch: your data objects, files, and the rest of the workspace.

People in the loop

Automation doesn't mean removing people from decisions that need them. A workflow can include a human review step that pauses the run and waits for someone to approve or reject before it continues. That's how a sensitive action stays automated in the routine parts while keeping a human gate exactly where you want one.

A lifecycle, not a switch

A workflow isn't live the moment you create it. It starts as a draft you can build and check, and you activate it when it's ready to run. That separation means you can design a process safely before anything actually fires.

Workflows and the Agent

The two automation tools in Spark are complementary. Use a workflow when the steps are known and you want a guarantee that they run the same way. Use the Agent when the work needs reasoning and the next step depends on what it finds. Many real processes use both: a workflow for the deterministic spine, the Agent for the judgment in between.

The Workflows section covers how to build, run, and debug them.