Reported operating use · Telecommunications
AI network diagnosis: One New Zealand example
One New Zealand uses a Google Cloud and Gemini deployment, with Future Connections, to investigate VoLTE voice-network issues. The supplier reports that most action paths keep human review. Fully closed-loop network operation is not established by the sources.
Voice-network diagnosis and reviewed remediation
One New Zealand
Operational VoLTE investigation deployment with human review on most action paths; fully closed-loop operation not established.
A published third-party example, not a Nuvepro customer case or evidence of a partnership.
The workflow at a glance
One New Zealand reported work
Illustrative summary of the published account
Reported work components. Their arrangement does not imply a fixed execution order or simultaneous work.
- Reported deployment
VoLTE investigation
A supplier reports investigation work at One NZ.
- Human accountability
Reviewed action paths
Most VoLTE action paths retain human review.
- Evidence distinction
Portfolio example is not deployment proof
Detailed supplier workflow mapping does not establish the exact One NZ implementation.
What the source reports
The supplier reports a VoLTE investigation deployment at One NZ. Its later retrospective says most VoLTE action paths retain human review. One NZ describes live network optimization and human accountability. The supplier's detailed L3 workflow-mapping example is portfolio practice, not direct evidence of the exact One NZ deployment.
Reported technology: Google Cloud / Gemini with Future Connections. Scope and outcomes are source-reported.
What the enterprise environment would need
The following is a proposed project design for an enterprise doing similar work. Published details are marked; the remaining systems, data and scenarios would be agreed for your project.
Proposed practice environment
Prepare a supported VoLTE diagnosis for review
Illustrative project design to scope with your enterprise.
- 1 · Set up the environment
VoLTE trace replay
Approved fictional telemetry, equipment configuration and counter definitions share aligned timestamps.
- 2 · Preparation and agent contributions
Prepare diagnosis and remediation
Identify supported network causes and draft a bounded engineering handoff.
- 3 · Independent checks
Check incident evidence independently
Known outcomes test counter semantics, timing and misleading correlations.
- 4 · People decide acceptance
Network engineer accepts recommendation
Review change authority and rollback before accepting practice findings.
A reviewable evidence package
A telemetry-supported diagnosis, uncertainty log and remediation draft.
If a check fails → withhold acceptance.
Misaligned counters create a convincing but unsupported diagnosis. Withhold acceptance, repair, independently recheck and return to owner review.
Failure or owner rejection returns to review
- Hold
- → 2 · Revise work
- → 3 · Rerun independent checks
- → 4 · Owner review
Operating boundary. Read-only traces or controlled testbed only; no live network changes or closed-loop claim.
- Systems and tools
- Inferred test scope: Existing network testbed or read-only replay of approved telemetry.
- Data and state
- Inferred test scope: Aligned timestamps, equipment configuration and counter semantics.
- People and handoffs
- Inferred test scope: Engineer reviews proposed remediation and owns operating impact.
- Rules and authority
- Inferred test scope: Allowed network changes require explicit authority and safe rollback.
- Normal and exception cases
- Proposed test cases: Actual outage, stale counter, conflicting telemetry, tool failure, false correlation.
- Reset and replay
- Not reported. Proposed: Replay historical traces; reset only a controlled testbed, never live network state.
A proposed simulation
An illustrative practice design derived from this work, not an executed assessment.
Diagnose a VoLTE issue amid stale counters and misleading correlations, then hand supported remediation to an authorized engineer.
Review gate: Check counter semantics, known incident cause, confidence, change authority and rollback; do not treat a fluent diagnosis as correct.
Challenge the result: Misalign counter timestamps or inject a misleading correlation. The assessment must detect unsupported diagnosis and prevent an unauthorized change.
What needs checking
- Published supplier practice: consistent tool schemas, structured errors, safe retries and provenance.
- Proposed: test diagnosis and recommendations against incident outcomes and engineer review.
- Published supplier lesson: explicit workflow mapping precedes agent design.
- Proposed: validate telemetry semantics and temporal alignment against operational reference traces.
Review environment fidelity, human capability and agent performance separately. The enterprise’s accountable owners decide acceptance of the shared workflow and any operating use.
Questions about this example
- What does One New Zealand use AI for?
- The supplier reports a VoLTE investigation deployment at One NZ using Google Cloud and Gemini with Future Connections. The work is voice-network diagnosis and reviewed remediation. One NZ also describes live network optimization with human accountability.
- Where do people stay involved in One NZ's network work?
- The supplier's retrospective says most VoLTE action paths retain human review, and One NZ describes human accountability. A proposed design, not One NZ's, has an engineer review the recommended fix and own its operating impact.
- How strong is the evidence for the One NZ example?
- It is reported use, from the supplier and One NZ. The supplier's detailed workflow-mapping example is portfolio practice, not direct evidence of this deployment. Independent incident-outcome accuracy and evidence for unrestricted closed-loop network operation are not supplied.
- What would an enterprise need to build something similar?
- This is Nuvepro's proposed design, not One NZ's. It would need a network testbed or read-only replay of approved telemetry, aligned timestamps and counter definitions. Tests would include stale counters and misleading correlations, with changes limited to authorized engineers and rollback.
Shape this around your enterprise
Bring a similar workflow and the people responsible for it. Together, define the records and systems to use, the result your team will build or configure, and how its owners will check it. Nuvepro scopes the simulation, environment and preparation. Agree any existing-agent evaluation, implementation assistance and production responsibilities separately.
Sources and limitations
Curation review: . This is the date this source example was reviewed, separate from source publication dates.
- Independent incident-outcome accuracy, full reset design and evidence for unrestricted closed-loop network operation are not supplied.
- Scaling the autonomous network: introducing the Data Steward & Core Network agentsSource publication: 2026-03-04
- A year of building agents for telco operatorsSource publication: 2026-05-26
- Responsible Artificial Intelligence (AI)
More it operations examples
- AI cloud incident investigation: Eon exampleEon
- AI disaster recovery evidence: Nubank exampleNubank
- AI incident investigation: InEight exampleInEight
- AI IT service requests: Redpanda exampleUnnamed enterprise customer reported by Redpanda