Harnessing Generative AI for Telcos: Boosting OSS/BSS Performance and Enabling Data-Sharing Services

The wave of popular enthusiasm surrounding generative Artificial Intelligence (AI) is often accompanied with vague claims for market disruption. The telecoms industry has also been enthused, but has shown recent progress in unveiling practical use cases for generative AI, with Amazon Web Services (AWS) reporting 19% of telcos globally committed to some use case. Telcos’ primary interests have integrating generative AI with Operations Support System (OSS)/Business Support System (BSS) functions for 1) enhancing operational efficiency or 2) creating entirely new opportunities for monetization.

On the side of boosting operational efficiency, Netcracker’s GenAI Telco solution is informative. First, with access to live telco data trained to generate language, speech, images, and code, generative AI models provide fluent and efficient information for consumers (Business-to-Consumer (B2C)) or business customers (Business-to-Business (B2B)), improving customer service and making faster sales. Second, generative AI is used for internal business operations, including analysis and marketing, and can provide recommendations for new telco service offerings. And finally, generative AI can support network operations and analytics, both in prompt-driven network design and automation, improving call center efficiency.

On the other hand, generative AI can support entirely new avenues for telcos to monetize customer data. For instance, Vodafone is using generative AI to generate Customer Data Records (CDRs) through synthetic data. By creating new data that imitates real data in the patterning of all relevant features, synthetic data enable telcos to share data with third parties without the risks of re-identification that have long troubled data-sharing. An entire ecosystem is emerging surrounding synthetic data generation with telcos as a strategic vertical; YData in the United States, CloudTDMS in London, and Betterdata in Singapore are already synthesizing telco network data.

For most telcos, generative AI is still at the Proof of Concept (PoC) stage, but its implementation in cases such as those highlighted do prove several qualities essential for market impact: relevance, versatility, security, and monetizability.

• Relevance: Training generative AI models on telco data promotes relevant language and code output, which can then be streamlined into business or network operations without issues related to translation or compatibility. This quality is well represented within Red Hat and IBM’s collaborative product, Ansible Lightspeed with Watson Code Assistant, which can automate network code and overcome a problematic DevOps skills gap.
• Versatility: In the age of cloud-native, hybrid, and multi-cloud, the OSS/BSS workload is more mobile than ever. Native integration of generative AI with the OSS/BSS workload upon which it is dependent permits deployment across public and private cloud environments.
• Security: Sensitive OSS/BSS data must not leak into public generative AI models or to generative AI users (especially consumers, business customers, and partners). Isolating and encrypting customer data and allowing access only through telco-grade secure gateways is an essential practice, and Netcracker’s solution provides a significant PoC here.
• Monetizability: In addition to enhancing operational efficiency, using generative AI to synthesize data for compliant data sharing allows telcos to tap into new revenue streams, partnering with data analytics firms and research institutions, cybersecurity companies, governments and urban planners, and other external parties interested in telco data.

By generating new content from OSS/BSS data patterns, generative AI provides unique solutions beyond the reach of discriminative AI, which is typically limited to classification tasks. Considering that generative and discriminative AI may also work together for 5G Core (5GC) network analytics, such as anomaly detection, ABI Research strongly expects generative AI to be integrated into both long-term and short-term telco strategies.

In the long term, we expect telcos to invest in bolstering automation with generative AI, be it in the OSS/BSS workload or the 5GC network. A long-term vision is appropriate for cautiously addressing the risks and costs associated with 1) securing exchange among customer data, generative AI models, and users; and 2) building telco-specific knowledge bases that are used to enrich generative AI with a company’s own information and practices to ensure quality output.  From the network vendors’ perspective, demands for telco-specific generative AI models provide new product opportunities; vendors’ services are also required for overcoming the operational challenges of implementing generative AI within their existing stacks.

On the other hand, telcos may promptly implement generative AI for synthetic data generation. For telcos already granting some access to data, it is strictly risk-minimizing. For those telcos that do not yet share data, synthetic data sharing offers a new revenue stream, while also addressing compliance and strict data-privacy standards.