# Telecommunications > Join network, subscriber, mobile money and partner data into one model, and give operators a private place to run AI inference close to their customers. Vertical ontology, workflows and compliance notes for telecommunications in Africa. Canonical URL: https://www.boltcliq.com/industries/telecommunications Part of: https://www.boltcliq.com/llms.txt Join network, subscriber, mobile money and partner data into one model, and give operators a private place to run AI inference close to their customers. - Sector pack: Starting ontology for networks, subscribers, wallets and partners - OSS + BSS: Typical sources: network, billing, CRM, mobile money and agent systems - Licence + DPA: Regulators in scope: telecom authority and data protection - USSD, SMS, app: Channels: USSD, SMS, WhatsApp, retail and agent networks ### The challenge - African operators run some of the largest data estates in their countries. Network systems record cell and tower performance. Billing and CRM hold subscriber plans and usage. Mobile money platforms hold wallets and transactions, often operated as a separate business unit. Dealers and agents sell airtime and SIMs. Tower companies, fiber partners and vendors keep their own records. USSD, SMS and WhatsApp are the main way customers reach the operator. - Each of these systems answers a narrow question. Network teams see congestion but not which valuable customers it hurts. Marketing sees churn risk but not the tower outages that caused it. Fraud teams see SIM swap patterns only after a wallet is drained. Operators want to run AI close to the data, but GPUs are scarce and expensive, cloud costs are priced in dollars, and customer data cannot always leave the country. Without a common model and strict access control, every AI pilot becomes a one off integration. ### The ontology **Objects** - 13 core things - The nouns of the sector. Each object has an identity, properties and a history, and maps to records in the source systems. - Subscriber - SIM - Device - Plan - Wallet - Cell site - Tower - Network incident - Ticket - Dealer - Agent - Campaign - Partner **Links** - 8 core relationships - How the objects connect. Links turn separate tables into a graph that people and agents can walk. - Subscriber holds SIM and Plan - SIM is used in Device - Subscriber owns Wallet - SIM attaches to Cell site - Cell site is hosted on Tower - Network incident affects Cell sites and Subscribers - Ticket is raised by Subscriber about Incident - Dealer and Agent sell to Subscriber **Actions** - 6 governed actions - What people and approved agents may do. Each action has rules, an approver where needed and a log entry. - Open and prioritize a network incident - Dispatch a field engineer to a site - Block or review a SIM swap - Send a targeted offer or apology - Escalate a customer ticket - Rebalance agent float or airtime stock ### Use cases - Network incidents - A set of towers degrade in a district during a power outage. Customers complain on social media before the network operations center confirms the scope. - Cell site alarms, power data and subscriber impact link in one incident. The system ranks sites by affected subscribers and valuable accounts, and suggests which engineer to dispatch. - The network operations lead sets priorities and approves field dispatch. Engineers confirm the cause. - Churn and care - A prepaid subscriber leaves after weeks of dropped calls. The retention team sends a generic promotion the following month. - The subscriber object holds usage, complaints and network experience. The system spots at risk customers and drafts a tailored message, such as a data bonus with an apology for known outages. - Marketing sets offers and budgets. Customers can opt out, and consent rules are applied by the system. - SIM swap fraud - Fraudsters port or swap a SIM, then drain the mobile money wallet within an hour. Telecom and money teams see different parts of the event. - SIM events, device changes and wallet activity are joined in near real time. Suspicious sequences raise an alert with evidence, and policy can require a hold or a call back. - Fraud analysts release holds and decide on escalation. Thresholds are set by risk owners. - Agent and dealer distribution - Agents run out of airtime or float in busy markets. Field sales teams visit outlets on fixed routes. - Sales, stock and location data feed route suggestions and restock alerts. Sales teams get a daily plan on a phone app or by SMS. - Regional managers adjust territories and targets. Agents decide what to stock. ### Data you already have - We read from your existing systems first and write back only through approved actions. Every source has a named owner. Where connectivity is weak or records are on paper, we plan for queued sync and OCR with human review. - Network management and performance systems - Billing, rating and charging platforms - CRM, care and ticketing tools - Mobile money platform and wallet ledgers - Dealer, agent and distribution systems - Tower, power and field service records - USSD, SMS and WhatsApp interaction logs - Device and SIM inventory databases ### Compliance and residency - Nigeria - NDPA 2023 and telecom authority rules - Subscriber and traffic data are personal data. Consent and retention rules shape analytics and marketing. - Kenya - Data Protection Act 2019 - Profiling and direct marketing need a lawful basis, and data subjects have rights over automated processing. - South Africa - POPIA - Direct marketing and automated decision rules may affect churn and offer models. - Ghana - Data Protection Act 2012 - Controllers and processors follow security and purpose principles for subscriber data. - Egypt - Personal Data Protection Law 2020 - Rules on consent and cross-border transfer may affect shared AI infrastructure. - Licence terms, SIM registration, lawful interception and mobile money rules vary by country and regulator. This list is illustrative. Confirm with counsel before designing data flows. Examples only, not legal advice. Confirm with counsel. ### Boltcliq products **Zeus** - Model plane - Models subscribers, SIMs, cell sites, incidents and wallets, so network, care, marketing and fraud teams see one customer experience. **Odin** - Trust plane - Controls internal, partner and tenant access to subscriber data, with purpose limits, so a vendor or enterprise customer sees only its slice. **Monolith** - Control plane - Provides the gateway for models and tools, with routing across large, small and local models, redaction and cost caps. **Argus** - Runtime plane - Runs agents and data near the operator, on local capacity or AWS, with portability for in-country inference. **Cyclopes** - Observation plane - Observes agents, models and data access, including cost per request, so operators can attribute spend and show how a decision was made. ### Frequently asked questions - Is Boltcliq a network analytics vendor? - No. We provide the ontology, governance and runtime layers. Operators and their vendors keep network tooling. - Can we run AI inside our own country? - That is a design goal. Argus is portable, and the model gateway can route to local models and capacity where the operator has it. - How is subscriber privacy protected? - Access is role and purpose based, sensitive fields can be masked from agents, and use is logged. Consent rules are enforced in workflows. - Does this help the mobile money unit? - Yes. Wallet and agent data join the subscriber model, which supports fraud checks and agent management, subject to separate entity rules. - Do you offer GPUs? - Not as a product today. Our neo cloud and inference work is a separate solution that depends on local partners.