Checked against Atlassian guidance on Sep 6, 2026

A practical Jira Automation step audit checklist.

Use export analysis to decide where to investigate, then use runtime evidence to determine what actually consumed steps. Static JSON alone cannot establish an exact monthly bill.

1

Organization usage

Confirm that usage is material

In Atlassian Administration, open Insights, then Platform usage. Compare current automation steps with the organization-level allowance.

2

Export JSON

Export the automation flows

From Jira settings, open System, then Automation flows. Use More actions and Export flows. Jira downloads global and project-scoped flows in one JSON file.

3

Static structure

Rank structural intensity

Count configured triggers, conditions, actions, branches, and loops. Prioritize frequent rules, broad scheduled queries, and flows with branch or loop fan-out.

4

Run logs

Inspect runtime evidence

Use run logs to check execution frequency, early exits, NO_MATCH triggers, failures, and the number of items processed by branches or loops.

5

Manual review

Validate the proposed change

Treat narrower queries and removed duplicate conditions as candidates. An admin must confirm that the behavior remains equivalent before importing a rewrite.

6

Before and after

Measure after the change

Import changes disabled, verify scope and owners, then enable deliberately. Compare later usage and run-log behavior with the baseline.

Worked structural example

One trigger, one branch, one condition, one action.

TriggerBranchConditionAction

The represented structure contains four known metered components. Actual consumption depends on which components execute and how branch cardinality behaves at runtime.

{
  "name": "Update linked issue status",
  "trigger": { "component": "TRIGGER" },
  "components": [
    {
      "component": "BRANCH",
      "children": [
        { "component": "CONDITION" },
        { "component": "ACTION" }
      ]
    }
  ]
}

Interpretation rules

Keep configuration and execution separate.

  • A trigger that returns NO_MATCH still counts when it executes.
  • A condition that stops a flow prevents later components from executing.
  • NO_ACTIONS_PERFORMED, SOME_ERRORS, and ABORTED can still represent metered executed steps.
  • Branch paths and loop iterations require runtime evidence; do not infer them from structure alone.
  • Agentic steps such as Rovo agents use separate credits rather than automation steps.