With the European AI Act entering into force in 2024, telecom operators face a critical compliance conflict: they cannot legally or effectively deploy AI models due to severe data fragmentation across legacy systems.
Data challenges in telecom AI implementation
National telecom operators often face fragmented and inconsistent data when launching AI projects. This is driven by legacy OSS/BSS ecosystems, which can include 15 to 25 disparate systems. Without a single customer view and unified reference data, training AI models yields poor results.
What is at stake for the industry
The European AI Act imposes strict requirements on the transparency and auditability of intelligent systems in VoIP networks and contact centers. This makes it impossible to deploy voice bots or routing systems without proper data governance.
A common mistake among operators is focusing on the technical accuracy of AI models rather than clear business metrics like operational cost reduction or ROI. Consistent implementation of Data Governance practices helps minimize legacy integration risks and ensure regulatory compliance.
What to do next
To overcome architectural chaos and ensure compliance, telecom operators should adopt a systematic approach to Data Governance:
- Implement MDM (Master Data Management) using a consolidation model for data aggregation or a registry model with APIs for two-way synchronization.
- Adopt a Data Mesh approach where each business domain is independently responsible for its own data quality.
- Leverage modern technologies: use Apache Kafka for asynchronous event delivery, API Gateway to unify access to customer data, and Kubernetes to automate microservices scaling during peak loads.
Prepared by a Software Ukraine member. Original publication.