GenAI for Telecoms

This also fits against a complex backdrop of skill sets, cybersecurity considerations, data availability, energy demands, customer focus and local AI regulation. GenAI capabilities are being embedded in smartphones, laptops and home gateways.
A short article cannot cover all the angles and practicalities, but instead provides a high-level map of the diverse GenAI landscape giving telcos guidance on areas for deeper exploration.

Internal GenAI use-case domains for telcos

Operators can exploit GenAI for multiple purposes. Although there are ideas across the entire telecoms business, the main emphasis and opportunities currently lie in:

• Customer service and experience
• Network deployment and operations
• Back-office functions
• New service creation and platforms

Various horizontal capabilities of GenAI appear in multiple contexts. Text and image creation (including multilingual support), summarisation, synthetic data creation, software code generation, anomaly detection and ongoing learning are common. In addition, agents can orchestrate or chain multiple tasks from a single request.

GenAI and the customer experience.

Improving customer interactions is one of the most visible domains for GenAI in telcos. Lifecycle tasks often involve collection, analysis, insight and action based on text and spoken data. GenAI is ideal for assisting or automating:

• Self-care websites and apps
• Multi-lingual chatbots
• Improved classification of requests and problems
• Customer service agent tools, such as case summarisation or suggested replies
• Personalisation of offers and adverts, including customised imagery or video
• Churn prediction and customer retention management

GenAI and network engineering / operations

The network itself is another major area for GenAI, mostly for specialist internal users in planning, engineering or operations teams. Workflow improvements may often combine GenAI with other AI techniques, as well as (perhaps) telco-specific models. Examples include:

• Summarisation of logs, alarms and other network datasets
• Anomaly detection for security risks or preventative maintenance
• AI-native radio technology, for optimising beam-forming or modelling of local radio conditions.
• Improved network planning and design, based on telemetry from live networks, mapping and digital twins, plus synthetic data
• Disaster response, such as reconfiguration during emergencies or after infrastructure damage

These use-cases should enhance network performance, predictability and operational productivity. However, a cautious approach to testing and deployment is essential, keeping a “human in the loop”, until AI techniques prove themselves robust to unusual situations.

Back office functions

Telco back-office functions hold numerous opportunities for GenAI, often similar to those in other industries, Models capable of telco-specific ontology and vocabulary will be important, especially in specialist areas such as legal and regulatory drafting, or creating responses to RFPs.

This area may also be impacted by regulatory changes on AI, especially where it regards HR issues, procurement or regulatory compliance.

Enabling GenAI applications

Telcos should explore creation, enablement or hosting of GenAI applications for consumer or enterprise customers and third-party developers. Some operators are considering "AI factories", especially where there is a national role to play in delivering "sovereign AI" infrastructure or platforms. This can link to development of local-language LLMs, as well as partnering with governments in areas such as national identity registers, healthcare or education.

There is also potential for GenAI based services created inhouse or through telco consortia and alliances – perhaps in areas such as gaming, customised advertising or smart home services.

Operators should be more skeptical about offering AI-optimised compute and "GPU-aaS", which could reuse spare capacity either in the main datacentres, edge nodes for AI-RAN, or next-generation home gateways. While heavily publicised, it is not clear what the go-to-market is, nor which types of workloads are optimal to run inside a telco’s infrastructure.

Impact on network traffic

The impact of GenAI on network traffic levels is still unclear. At present, most GenAI applications perform inferencing on the cloud, transmitting relatively small amounts of data to the end-user, such as text or static images. Future applications such as GenAI-created video or AR/VR applications, could increase both uplink and downlink loads.

That said, AI might reduce traffic, if use displaces other activities such as watching video, or replaces extensive browsing with shorter AI summaries. Agentic AI could replace long, complex user sessions with orchestrated actions. Semantic compression of video could emerge, where AI encodes a scene into a text description and a few images, reducing gigabytes of data to megabytes.

Summary

GenAI is likely to be transformative. CTOs’ teams face a huge surface for efficiency gains, productivity enhancements and new revenue streams. But it is important to recognise limitations, infrastructure and skills needs and inevitable hype. Operators should also pay attention to wider trends around AI-capable user devices, national GenAI initiatives and regulation, as well as AI-RAN and AI-Native 6G development.

The low-hanging GenAI fruit is currently around customer service functions, back-office applications, plus certain network operations tasks. Less certain are platforms for customer-facing GenAI applications, and ideas for reselling AI edge compute.

Finally – for all its huge potential, it is also important to recognise that GenAI is just one part of a much wider AI landscape for operators.