Customer Case Study · Telecom · Caribbean
Two backbone platforms, one governed operations layer
How a national telecommunications provider in the Caribbean runs its retail operation on two mission-critical platforms, a customer-care console that unifies billing, network and VAS state across more than ten integrated systems and orchestrates IMSI change as one governed workflow, and a SIM lifecycle system that is the single source of truth from vendor order to activation, with Opstral as the operations layer above both: every integration, service and server under telemetry, Sentinel on every incident, and around ninety percent of L1 tasks automated under governance.
- 2mission-critical platforms, the backbone of the provider's retail operation, under one governed ops layerThe customer-care console and the SIM lifecycle system
- 10+systems integrated behind the care console: billing, 3G and 4G network cores, VAS platformsIntegrated in production
- ~90%of L1 operations tasks automated under governed proceduresAs operated in this deployment
- One viewbilling, network and VAS state for a subscriber, in a single consolePreviously one console per system
- One workflowIMSI change orchestrated across the 3G, 4G and billing systems it touchesPreviously a manual multi-system procedure
- End to endSIM lifecycle: vendor ordering, inventory, allocation and activation in one system of recordPreviously scattered records
The Challenges
The provider's retail operation depended on more than ten systems that had grown up separately: billing, 3G and 4G network cores, VAS platforms, provisioning. Customer care worked across all of them one console at a time; a routine IMSI change was a relay race across teams and systems; SIM inventory truth lived in fragments; and when any of the integrations holding this together degraded, the first alarm was usually a customer on the phone.
- Swivel-chair customer care
Billing, network and VAS state lived in separate consoles; answering one subscriber's question meant hopping between systems mid-call.
- IMSI change as a relay race
One SIM swap touched 3G, 4G and billing systems in sequence, manually, across teams, with every handoff a chance to get it wrong.
- SIM truth in fragments
Inventory, vendor ordering and activation state lived across records that did not agree; nobody had one answer to how many SIMs, where.
- Integrations nobody watched
Ten-plus integrations carried the operation, and a degraded one announced itself through call-center symptoms, not through monitoring.
- L1 buried in routine
The front line spent its day on repetitive checks and standard tasks, the volume of which scaled with subscribers, not with headcount.
- Cross-system RCA by hand
An incident spanning console, integration and backend meant teams comparing logs across systems until someone found the story.
The architecture, and why the ops layer is not optional
Two layers were delivered, and they are different kinds of thing. The applications, a customer-care console that unifies billing, network and VAS state per subscriber and orchestrates IMSI change as one governed workflow across the systems it touches, and a SIM lifecycle system that is the single source of truth from vendor order to activation, became the backbone of the provider's retail operation.
That is exactly why the second layer exists. Opstral sits above both platforms and everything they depend on: every integration, every service and every server under continuous telemetry, Sentinel investigating anomalies and incidents with probable cause and evidence attached, and governed automation carrying around ninety percent of L1 task volume. L2 and L3 engineers are augmented rather than replaced: they start from an evidence trail and an RCA, not from raw logs.
Governed autonomy throughout: L1 automation runs only inside the approved procedure set, every action is audited with rollback, and incidents carry their evidence. A platform this central to the business does not get discretionary automation, and should not.
What was switched on, and what was not
On a platform this central, the split below is the product. Around ninety percent of L1 task volume runs under approved, audited procedures; everything novel, cross-system or judgement-bearing reaches an engineer, carrying its evidence.
Runs without a human
- Telemetry collection across every integration, service and server behind both platforms, continuously
- Health of each of the ten-plus integrations, watched directly rather than inferred from complaints
- Sentinel investigation of anomalies and incidents, with probable cause and evidence attached
- L1 routine tasks in the approved procedure set, around ninety percent of L1 volume, audited with rollback
- The SIM lifecycle record: vendor order, inventory, allocation and activation state, kept as one truth
Does not run without a human
- Any action outside the approved procedure set, however confident the diagnosis
- Expanding the automation scope: every new procedure is written, reviewed and approved by people
- Novel and cross-system incidents: L2 and L3 conclude and act, starting from the evidence trail
- Changes to the orchestration workflows the business runs on, IMSI change included
What happens now when an integration degrades
The sequence below is what replaced finding out from the call center. It is the same loop whether the first sign is an error rate on a billing integration or a latency drift on a network-facing service.
- Telemetry notices the drift before an agent does
Error rates, latency and health signals stream from every integration, service and server. A degrading integration shows up as data minutes before it shows up as a queue of confused calls.
PlatformObserve - Sentinel investigates across the estate
The anomaly is correlated across the console, the integration and the backend system behind it. The output is a probable cause with its evidence, not a page to a human.
PlatformInvestigate - An incident is raised carrying the cause
Probable cause, evidence and the affected platform capability arrive together as the incident is created, so triage starts from a case rather than from raw logs.
PlatformInvestigate - A known pattern executes under governance
Where the fault matches an approved procedure, it runs as an audited, reversible action. This is the path that carries around ninety percent of L1 task volume.
PlatformAct - Everything else reaches L2 and L3, augmented
Novel and cross-system incidents go to engineers with the evidence trail and RCA attached. The engineers conclude and act; the platform did the archaeology.
Named engineerAct - Repeat offenders become data
Recurring causes accumulate across the estate rather than across anecdotes, so the procedure set and the integrations themselves keep improving.
PlatformOptimize
The Impact
Customer care answers from one console instead of five. A SIM swap is a workflow, not a relay race. And the platforms the business stands on are watched, investigated and largely operated by the platform underneath them, with people kept exactly where judgement lives.
| Dimension | Before | After · unified & governed |
|---|---|---|
| Customer care | One console per system; swivel-chair answers mid-call | Billing, network and VAS state in one console, per subscriber |
| IMSI change | Manual relay across 3G, 4G and billing systems and teams | One orchestrated, governed workflow across every system it touches |
| SIM inventory | Fragmented records; ordering to vendor driven by hand | One system of record from vendor order to activation |
| Integration health | Discovered through call-center symptoms | Every integration under telemetry, degradation surfaced as data |
| L1 operations | Front line consumed by repetitive routine tasks | ~90% of L1 task volume automated under approved, audited procedures |
| L2 / L3 resolution | Cross-system triage by comparing logs across teams | Engineers start from Sentinel's evidence trail and RCA, augmented rather than replaced |
These two platforms are the backbone of the retail operation, which is exactly why they could not be left to conventional monitoring. Putting one ops layer over every integration, service and server, with automation carrying the routine and engineers carrying the judgement, is what makes mission-critical sustainable.
At a glance
- Industry
- National Telecommunications Provider
- Geography
- Caribbean
- Engagement
- Production: applications delivered and operated
- Scope
- Customer-Care Console · SIM Lifecycle · Autonomous Operations
- Platform components
- Sentinel AI, full-estate telemetry, governed L1 automation, orchestrated workflows
- Evidence basis
- The ~90% L1 automation share is as operated in this deployment; qualitative improvements are described as such, without invented percentages
Put your backbone applications under one governed ops layer
Full-estate telemetry, Sentinel on every incident, and governed automation for the routine, deployed inside your environment, for the platforms your business cannot afford to operate on hope.
Want the full depth?
The extended PDF covers the logical architecture, the governance boundary, the engagement model and the delivery timeline, with the basis stated for every figure.
Customer identity withheld by request and referred to throughout as a national telecommunications provider in the Caribbean. The ~90% L1 automation share is as operated in this deployment; where no measured baseline exists, improvements are described qualitatively rather than as invented percentages. Exact results vary by estate and task mix.