AI that runs the work operators used to do by hand.
Not a chatbot bolted onto a billing system. Three operational AI capabilities — agents, anomaly detection, and stream monitoring — running continuously on top of the same Selcomm rating, billing and device management engine that has been in production for two decades.
Why AI here, why now
IoT operators are managing fleets that no human team can keep up with — millions of SIMs, meters and sensors, each generating telemetry on its own clock, each subject to its own tariff, each capable of failing in its own quiet way. The platform’s AI layer is built specifically for that reality.
It runs on the same data the rest of the platform sees. It does not need a separate data lake, a separate model team, or a six-month integration. It is part of the platform.
AI agents that orchestrate device fleets
Autonomous agents that manage device fleets and data collection across constantly-changing structures — provisioning, reshaping, and recovering devices without human ticket-by-ticket effort.
The first pillar is operational. IoT estates do not stand still — devices are constantly being added, retired, moved between tariffs, and re-grouped under new commercial structures. Doing that work ticket-by-ticket is what makes IoT operations expensive, slow and brittle.
The orchestration agents take that work on directly. They watch the device fleet, the rating engine, the CRM and the network telemetry simultaneously, and act on the same configuration data your operators do — so anything an agent does is auditable, reversible, and fully consistent with how a human would have done it.
What’s different is the volume and the speed. An agent can reshape ten thousand devices into a new billing structure in the time it takes to write the ticket.
What it does
- Continuously discover and onboard new devices into the right billing and management group
- Reshape fleets when networks, tariffs or operational structures change
- Detect and remediate stuck, silent or misconfigured devices automatically
- Coordinate bulk provisioning, decommissioning and firmware operations
- Hand off to humans only for genuinely novel exceptions
Anomaly & revenue-leakage detection
AI watches every device, every tariff and every usage stream for anomalies that quietly become revenue leakage — and surfaces them while there's still time to act.
The second pillar is the most quietly valuable. In any large IoT estate, a small percentage of devices are mis-rated, mis-grouped, mis-provisioned or silently failing — and that small percentage compounds into material revenue leakage every billing cycle.
The anomaly engine learns each device’s normal behaviour against the rest of the fleet, against historical baselines, and against the tariff configuration in place. When the picture stops making sense, the platform flags it — with enough context that a human can decide what to do, or an orchestration agent can fix it without one.
What it does
- Per-device usage profiles that learn what 'normal' looks like
- Cross-tariff comparisons to surface mis-rated services
- Silent-device detection — meters that should be reporting but aren't
- Configuration drift alerts when a device's billing setup diverges from its peers
- Closed-loop remediation — open the exception, fix it, reconcile the invoice
Real-time billing-stream monitoring
Every billing stream — rated event, invoice line, payment — is watched continuously. Issues are caught the moment they appear, not the morning after the run.
The third pillar is about closing the time-to-detect gap. Traditional billing assurance runs after the cycle — by which point the invoice has already gone out, the customer has already noticed, and the credit memo is already being written.
The billing-stream monitor watches every rated event as it lands, every invoice line as it forms, and every payment as it clears. When a stream behaves in a way the platform doesn’t recognise, it’s surfaced immediately — not a week later in a reconciliation report.
What it does
- Real-time inspection of rated events as they're produced
- Pattern matching against known issue signatures
- Automatic correlation between device events and billing outcomes
- Continuous reconciliation between provisioning, telemetry and invoicing
- Operator-friendly dashboards plus webhook and API hooks
Want to put one of these to work?
Pick the pillar that maps to your highest-value problem. We'll come back with a concrete pilot scope.