10 billion CDRs, one question that never gets answered in time: Ask your network in Arabic, get the SQL in 4 seconds
NOC engineers, revenue-assurance leads and CVM marketers ask BSS, OSS, CDR and CRM data in natural language. CST-aware controls, PDPL-grade PII masking on MSISDN/IMSI/IMEI, per-region row-level security, and a semantic layer that finally reconciles ARPU, churn and revenue leakage across every source.
- NL2SQL accuracy
- 94.6%NL2SQL accuracy
- on 10B-row CDR joins
- < 4son 10B-row CDR joins
- on telecom billing eval
- 87%on telecom billing eval
- reconciled everywhere
- 6 KPIsreconciled everywhere
In Saudi Arabia
What a telecom reviewer checks before an NL2SQL bot goes live
Every question, Arabic or English, runs through the same governed semantic layer. This is the evidence the sector authority expects before the bot touches production numbers.
CST
Metric definitions used by the bot match the definitions الاتصالات reporting already uses.
Semantic layer built from the existing governed metric catalogue, not invented per question.
Metric-to-source mapping document reviewed by the reporting owner.
SDAIA, PDPL
Row and column-level entitlements applied before an Arabic or English question can return a number.
Identity resolved and entitlements applied at the semantic layer, ahead of query generation.
Entitlement test log showing the same question answered differently per role.
NCA ECC
Read-only execution path, with SQL validated and cost-checked before it runs.
Query validator gate ahead of execution, no write path exposed to the bot.
SQL audit trail attached to every answer, reviewable line by line.
Where it runs
The warehouse connection and the semantic layer run inside Saudi Arabia; the bot never ships raw data outside that boundary to generate an answer.
Arabic, as it is spoken here
Arabic business questions, including mixed Arabic-English phrasing common in Saudi teams, resolve against the same catalogue without a translation step.
Free 30-day PoC on a sanitized slice of your CDR mediation + BSS
87%+ accuracy validated on industry telecom NL2SQL benchmark
Runs on your Iceberg / Delta lakehouse, no re-platforming
Arabic-first CVM prompts: Saudi register, not translated MSA
Questions your team actually asks, answered with governed SQL
In Arabic, English or a mix, the bot converts each question into SQL over a unified semantic layer.
5G sites in Riyadh with PRB utilization above 80% during peak hour yesterday.
أعلى 50 مشتركاً من حيث تراجع ARPU خلال آخر 30 يوماً في المنطقة الشرقية
Voluntary churn rate by prepaid vs postpaid, split by dealer channel, last quarter.
أي مواقع سجّلت انقطاعات أكثر من مرتين هذا الأسبوع مع تأثير SLA؟
Revenue leakage flags on interconnect billing this week, ranked by SAR impact.
أعرض حملات CVM التي سجّلت أعلى ROI في جدة الشهر الماضي
Where teams use NL2SQL in nl2sql for telecom operators
NOC engineer, 3am incident
Engineer asks the bot on WhatsApp: 'أي خلايا سجّلت drop calls فوق 3% في آخر ساعة؟'، governed SQL runs on the OSS lakehouse, PII stays masked, incident is triaged before the on-call escalates.
Revenue assurance, 7-figure leak found in a day
RA teams ask about interconnect leakage, roaming settlement gaps and B2B billing anomalies, every answer traces back to the reconciled semantic KPI and the raw rows, so finance can act, not just alert.
CVM marketing without a data-team ticket
Marketing asks 'لماذا ارتفع تخارج الدفع المسبق في الدمام الأسبوع الماضي؟'، the bot builds cohort curves, correlates with dealer performance, and returns Arabic explanations. Campaigns launch same-day, not same-quarter.
Certified semantic layer + complete enterprise security layer
We never wire the bot straight to your warehouse, every question runs through a governed semantic layer and a security stack that respects every row and column.
Certified semantic layer
- One definition for ARPU, MOU, blended churn, NPS, revenue leakage and interconnect margin.
- Semantic layer over dbt / Cube covering BSS, OSS, CDR mediation, dealer ledger and CRM.
- Streaming ingestion (Kafka + Iceberg) so operational questions answer within seconds, not overnight.
- Graceful failure: 'I can't reconcile this ARPU definition, did you mean blended or postpaid-only?'
Complete enterprise security layer
- PDPL-grade PII masking on MSISDN, IMSI, IMEI, national ID and email, role-based reveal.
- Row-level security per region, brand, dealer, business unit and time window.
- Read-only SQL sandbox, the bot cannot write to network or billing tables. Ever.
- Prompt-injection filters tuned on Arabic + Arabizi + code-switched CVM prompts.
- Immutable audit, every question, SQL, table, row-count and answer retained for CST review.
Questions Saudi leaders ask before rollout
Our CDR joins are 10B rows. Can it really answer in seconds?
Yes, on partitioned Iceberg or Delta tables, typical operational questions return in under 4 seconds. Heavy trend queries return within 10 seconds and are cached at the semantic layer for zero-cost re-asks. We prove this on YOUR data in the PoC before you commit.
Can the bot join CDR data with CRM without leaking PII?
All joins run behind the semantic layer with tokenized MSISDN/IMSI. PII is masked or revealed strictly per role. An analyst sees aggregated cohorts; a fraud investigator with the right role sees the raw identifier, and every reveal is logged for CST review.
We already have a BI-vendor Copilot on our warehouse, why add another tool?
Bundled Copilots are trained on English-labeled generic schemas and score in the 10–31% range on real telecom mediation (public benchmarks). Ours is Arabic-first, tuned on Saudi telecom terminology, runs in-country, and publishes accuracy on your sanitized data before you sign anything.
What about Gulf Arabic and code-switching? Our CVM team writes half-Arabic half-English.
The bot handles Saudi MSA, Najdi, Hejazi and Eastern-region variants, plus Arabizi (Latin-script Arabic) and mixed Arabic/English in the same sentence. Every dialect is measured in the published NL2SQL evaluation harness.
See your first network / CDR question answered on YOUR sanitized data, 20-minute working session, no PoC fees.
No fees, no procurement commitment, no data leaving Saudi Arabia. You walk out of the session with an accuracy evaluation report your team owns, whether you buy or not.