Operational incident response in Pūnaha Voice: en-NZ-MollyNeural (natural synthetic voice) Prepared Pūnaha walkthrough. Fictional service, incident and data. No customer result is implied. Nothing is sent or saved. Chapter 1: Open the incident case The journey starts at eight twelve in the morning. Koru Access, a fictional internal business service, is producing intermittent sign-in failures. The service owner opens one operational incident case and records what is known: the affected service, the symptoms, the operating window, the estimated reach and whether a workaround exists. Try changing the impact. That detail will follow the incident into the response-level decision. The case gives changing work a stable home. New evidence and tasks can be added without losing the original report or pretending that the first description was complete. Chapter 2: Gather approved context The incident case starts a published triage workflow. In this prepared example, the workflow reads a named service-status connection, retrieves the approved recovery runbook and finds two similar resolved incidents. A configured AI model can prepare a source-aware summary from that approved material. The sources and gaps remain visible, and a human reviewer confirms the incident picture. Select Gather approved context to reveal the prepared pack. The public page does not contact another system or model. It shows how repeatable triage could give the response team a useful starting point without turning a generated summary into an operational decision. Chapter 3: Explain the response level Now the incident lead applies a configurable response-level decision. The example looks at business reach, service criticality, the available workaround and confidence in the evidence. Change an answer and the result updates with the score, matched rule and reasons. For the prepared incident, several departments are affected, the service is operationally important and no reliable workaround has been confirmed. That selects a major response and requires incident lead approval. This is not a universal incident-severity method. Your organisation would set its own inputs, thresholds and authority. The important point is that the result is explainable and stays with the incident. Chapter 4: Activate the case plan The decision result enables the work that is required now. The adaptive case plan contains a containment workflow, a stakeholder update, a recovery milestone and a discretionary dependency review. Select Activate response plan. The required activities become active, while dependency review remains available for evidence that has not arrived yet. This is where an adaptive case differs from a rigid checklist. It can preserve the possible work without forcing every incident through every task. Conditions and events make the relevant work available as the situation develops, and the case continues to hold the shared context. Chapter 5: Start repeatable response work The case starts a mapped BPMN-based response workflow. This handles the work that should be coordinated consistently: preparing containment options, assigning specialists, reviewing a stakeholder update, waiting for human approval and then waiting for a recovery signal. Start the response workflow to see the linked run become active. The case and workflow have different jobs, but they remain connected. The case holds the changing incident. The process manages the agreed route and can wait without losing its place. In a real configuration, the mapped workflow version, roles, connections and approval boundaries would be chosen and tested by the organisation. Chapter 6: Adapt to new evidence At eight thirty-four, a new signal suggests that an identity dependency is experiencing unusual latency. Add that signal to the case. The evidence timeline retains the original report, the retrieved material and the earlier response decision. It now also makes dependency review available to an identity-platform specialist. The new evidence does not erase the earlier understanding, and the team does not need to restart the incident. The case changes the available work while preserving why the previous route was selected. This walkthrough uses prepared fictional evidence, but it demonstrates the practical value of combining evidence, sentries and discretionary work. Chapter 7: Keep a person responsible The incident lead receives a compact decision pack with the current finding, the response rule, the evidence and a prepared stakeholder update. Pūnaha can organise those inputs, but it does not approve a consequential action on the person's behalf. Choose controlled reroute. The walkthrough records the human decision and shows that the prepared communication is approved for a configured continuation. A real connection could pass approved fields to an incident service or email route, subject to the organisation's controls. This public demonstration sends nothing. It keeps the accountable decision visible and makes the boundary between preparation, approval and action clear. Chapter 8: Monitor recovery After the containment route is approved, the response workflow waits for a correlated recovery signal. This durable wait represents a thirty-minute monitoring window. Select Simulate stable monitoring result. The successful sign-in rate rises, dependency latency returns to its expected range and the recovery signal resumes the workflow. The service-recovered milestone is reached only after that configured evidence arrives. The public page is not connected to monitoring, so the values are prepared and fictional. The design point is that recovery is supported by visible evidence and a named event, rather than being declared simply because a task was completed. Chapter 9: Open the post-incident review Recovery is not the end of the work. Move the fictional clock forward forty-eight hours. A timer activates the post-incident review task and assigns it to the service-improvement lead. The task begins with the report, retrieved sources, case events, response decision, human approval and monitoring result already attached. The reviewer can examine what changed, update the runbook, begin problem analysis and assign follow-up actions without reconstructing the incident from separate messages. The full journey has combined an adaptive case, repeatable workflows, an explicit decision and human authority while keeping one inspectable history from report to learning.