Quality Assurance
Part of Automation governance
Defining a safe process for disabling an automation
Set stop authority, verify that automated action has ceased, classify affected items and reconcile outcomes before restart.
A safe disable process prevents further automated action, establishes what may already have happened and assigns every affected item a next decision. Write the stop authority, platform steps and restart conditions before an urgent failure occurs. Turning off a schedule or trigger alone does not prove that work has stopped.
Safe Process for Disabling an Automation
- Define stop authority and scopeSpecify who can stop the automation, under what conditions, and the full scope (e.g., case type, environment). Record time, reason, actor and version.
- Confirm the stop took effectVerify job and queue states across platforms. Check if new work can still start. Collect business references for running, queued and exception items.
- Control work during the pauseBusiness owner decides which cases wait and which continue manually. Maintain a controlled list of held items and manual completions with clear reference to automation attempts.
- Authorise restart from reconciled evidenceConfirm fault is resolved, reconcile held items and manual actions by business reference. Resume intake only after verifying destination outcomes and obtaining approval.
Define authority and scope
State when a planned pause is appropriate and when a technical responder may make an urgent stop. Name who orders the stop, who performs each platform action and who receives notice. An urgent containment action should be followed by business-owner review of affected work.
Specify whether the stop covers a case type, trigger, environment or whole automation. Identify every path that can start or continue work, including queued jobs, retries and connected workflows where they exist. Record the time, reason, actor and running version. Do not describe a partial stop as complete while another route remains active.
Confirm what the stop did
The technical owner should check the platform’s actual job and queue states and whether new work can still start. A stop request, disabled trigger or cancelled job may have different effects in different products and configurations. It may leave an action in progress or a delayed attempt pending. Record which effects are confirmed and which remain unknown.
Collect the business references for running, queued and exception items. A “last good item” is useful only for a strictly ordered process; parallel work needs item-level evidence. Check the destination before assuming an attempted write was reversed.
| Item state | Next decision |
|---|---|
| Confirmed complete in the destination | Retain evidence and prevent replay |
| Confirmed not started | Hold or assign an authorised route |
| Partly completed | Assess remaining effects before any continuation |
| Attempted action with unknown result | Hold until the destination can be checked or an authorised decision is made |
| Waiting for approval | Preserve that requirement during any manual route |
A timeout after submission leaves the result unknown until it is checked. If the destination cannot establish the state, keep the item with a named decision maker instead of treating a retry as safe.
Control work during the pause
The business owner decides which cases may wait and which may continue manually under existing approval rules. Tell affected teams where new work goes.
Keep a controlled list of held items and manual completions, joined by business reference to automation attempts and destination results. Detailed staffing and step-by-step manual processing for a stopped bot belong to the separate bot fallback procedure.
Authorise restart from reconciled evidence
Before restart, the technical owner confirms the fault or planned change has been addressed and the proposed version is ready. The business owner decides the treatment of open cases.
Compare held items, manual actions and destination records by business reference; remove or mark manually completed items through the approved platform method. Define what evidence permits intake to resume and who authorises it.
Where the process allows, begin with a controlled set of eligible items and check destination outcomes before restoring normal intake. Keep the stop record, item dispositions and restart decision together.
