Workflows

Create a workflow

Open Automations → Workflows and choose the create action. Give the workflow a name containing the business event and result. In the editor, configure the trigger first, then add actions or decision paths in the order data becomes available.

Exact nodes depend on workspace capabilities. Treat every node as a contract: it receives input, applies configuration, produces output, and can fail.

Configure triggers narrowly

Define the record type, event, and conditions precisely. Consider existing records, repeated updates, imports, merges, timezone boundaries, and events created by another automation. When a broad trigger is unavoidable, add an early eligibility condition.

Use variables safely

Use the variable picker rather than manually typing tokens when possible. Verify that the selected value comes from the intended object and exists at that point in the graph. Define a fallback or branch for missing optional data. Never allow an empty value to become an invalid recipient, public message, or incorrect record update.

Actions, conditions, and waits

• Put irreversible or customer-facing actions after eligibility checks.
• Keep conditions mutually understandable and define what the default branch means.
• Account for timezone and record changes during a delay.
• Recheck critical conditions after a long wait.
• Avoid loops in which a workflow update recreates its own trigger.
• Design retries so they do not repeat a completed external action.

Validate and test

Use the editor validation panel and resolve every error. Then test representative records for normal, missing-data, ineligible, provider-failure, and duplicate-event cases. Verify both the execution and the downstream result.

Enable and monitor

Enable only after the test produces the expected records, messages, tasks, or provider changes. Use the Executions view to inspect status and open a specific execution for node-level details. Use Analytics for trend-level success and failure, then drill into examples.

When a run fails:

1. Open the original execution before retrying.
2. Identify the first failed node and its input.
3. Check provider health and authorization.
4. Correct the record or configuration.
5. Determine whether retry is safe and whether any earlier actions already completed.

Pause, edit, and retire

Disable new entry before major changes. Review in-flight executions and dependent reports. When retiring a workflow, preserve enough name and audit context to interpret historical executions, and remove or update upstream links that still expect it.