One of many revenue-and-trust use cases. This example runs full-population reconciliation across mediation, rating, and billing, explains every discrepancy, and hands finance a reviewed export to correct.
Proves these Lumina features
Operators lose an estimated 3 to 10% of revenue to leakage: billing errors, reconciliation gaps, and usage that is rated but never billed. Across the industry that is on the order of a hundred billion dollars a year.
Large operators process more than a billion call-detail records a day. Rule-based assurance samples, so it misses new leakage patterns, and discrepancies hide in the seams between usage, rating, provisioning, and the ledger for weeks before anyone notices.
Traditional approaches were not built for the scale and complexity of modern operations.
Rule-based RA checks a sample after the fact. Leakage that does not match an existing rule, or that lives between two systems, never gets flagged.
Usage, rating, provisioning, and the ledger are separate systems. Leakage hides in the joins, exactly where no single system is looking.
Every new bundle, promotion, and partner settlement creates a new leakage surface. The rules ship slower than the products do.
A flagged discrepancy with no traceable reasoning cannot survive a finance or regulatory review. Auditors need the why, not just the what.
Three layers of intelligence working together: reasoning agents, proactive detection, and multi-agent deliberation.
Agents reconcile across mediation, rating, and billing on the full population, not a sample, and explain each discrepancy in plain language with the query behind it. Your rating rules stay outside the model, owned and auditable, through Separation of Logic.
Every calculation is SQL you can verify. No black box.
Reconciling 1.24B CDRs against rated + billed records (cycle 2026-05)... Matched: 1,238,902,144 Rated but never billed: 41,209 events, ~$182,400 Provisioned but never rated: 3,118 sessions, ~$54,900 → Root: roaming partner P-DE bundle mapped to a retired rate plan → Each case carries its source rows and the rule that should have applied
ANOMALY: New bundle 'Unlimited-5G-EU' shows usage up 220% but rated revenue flat Benford analysis on rated amounts deviates (p=0.002) → Likely rating gap on the new bundle. Flagged in cycle 1, not cycle 5.
The Radar scans every cycle for emerging leakage signatures: a new plan whose usage and revenue diverge, a settlement that drifts, a provisioning path with no rating record. It surfaces the leak in the first cycle, not the fifth.
The Radar surfaces issues the operator didn't know to look for. Before they become incidents.
Revenue-assurance, finance, and provisioning agents deliberate on each leakage cluster and produce a reviewed correction package. The output is a governed export of flagged cases for human-approved correction. Lumina never writes back to the billing system of record.
The output is grounded in facts (SQL results), not hallucination. Every recommendation carries a full audit trail.
RA Agent: 41,209 rated-but-unbilled events on partner P-DE, recommend rebill Finance Agent: rebill window is 60 days, 2,140 events are out of window Provisioning Agent: root is a retired rate-plan mapping, fix the mapping or it recurs → Export: 39,069 in-window cases for approval + 1 mapping correction ticket. Human signs off before any change.
Lumina addresses the four strategic problems that hold operators back.
Reconcile the full CDR population every cycle instead of a sample, across every billing seam, without adding analysts.
Recover leakage in the cycle it happens, not weeks later. On a 3 to 10% leakage base, early detection is direct margin.
Cortex remembers every leakage pattern and root cause, so the same plan-mapping gap is caught instantly the next time it appears.
Every discrepancy carries its source rows and the rule that should have applied. Findings are a governed export for human approval, audit-ready by construction.
Specialized AI agents that power this workflow.
Revenue Assurance Analyst
Reconciles mediation, rating, and billing on the full population and explains each discrepancy.
Fraud & Leakage Reasoner
Separates deliberate fraud from process leakage so each case routes to the right team.
Sovereignty & Custody
Keeps billing and CDR data inside your environment while the reconciliation runs.
This is just one of many use cases
Revenue leakage & assurance is one example of how Lumina reasons on operational data. Across Telecom, every domain has use cases where AI agents can add value.
Not ready to commit? Stay up to date as we release new capabilities and industry-specific agents.
View all Telecom use cases