In this informative briefing, AWS explains how telecom operators in the MEA region can leverage the Generative AI breakthrough in AI/ML.
Generative AI marks an advancement in AI and machine learning (ML), and its emergence presents an opportunity for forward-thinking telecom executives to expand their networks, innovate business models, and adopt new technologies while controlling expenses to generate value for stakeholders.
Generative AI is enabled by the affordability of large-scale computing, the availability of extensive data, and innovative cloud-based architectures to monetize that data. Unlike traditional approaches focused on prediction, classification, or optimization, generative AI, powered by versatile pre-trained foundation models, can create content.
The proven value of foundation models, including large language models (LLMs), positions generative AI as a strategic opportunity for telecoms. Quick wins have already demonstrated their value, secured leadership commitment and motivated developer teams within telecom operators and CSPs to expand AI/ML usage into this new field.
Promising Telecoms Use Cases
Early engagements with telecoms and AWS ecosystem partners have demonstrated how generative AI can reduce revenue leakage and customer churn, improve cross-selling attach rates, decrease customer support resolution times, and streamline network operations while reducing troubleshooting time.
AWS has collaborated with CSPs and operators around the world to foster fast-paced experimentation to vector in on and single out successful use case candidates that deliver tangible business outcomes in the area of increasing developer productivity. Additional use cases center on improved support for knowledge workers to gain insight and process information in different domains, such as customer service, network operations and business processes for revenue assurance and billing accuracy. And in the area of end-user experience, generative AI was successfully deployed to create hyper-personalized experiences for smart home users and a new customer interaction through super-bundling of services through digital portals.
Presented at Digital Transformation World 2023 in Copenhagen, AWS partners such as Salesforce, Tech Mahindra and Snowflake and operators such as TELUS and Optus demonstrated the implementation of different use cases with AWS. At AWS re:Invent 2023 in Las Vegas, additional implementations and presentations from operators such as Liberty Latin America, Globe and Cox Communications underpin the innovation journey of the telecom sector to embrace generative AI to accelerate their digital transformation.
Enabling MEA Operators to Seize the Opportunity
CSPs and operators in the Middle East and Africa seek to accelerate their generative AI journey through the use of cloud, and AWS is actively working to democratize access to these technological capabilities in the region to put the tools and mechanisms for innovation right into the hands of developers.
As the first building block, these ambitions in the Middle East and Africa region are supported by AWS Global Cloud Infrastructure. Globally, AWS offers over 27 launched regions with 87 availability zones and more than 410 PoPs, complemented by 21 local zones, 28 Wavelength zones and 115 Direct Connect locations in 245 countries and territories. This includes the AWS Middle East (UAE) region and the AWS Middle East (Bahrain) region.
At the same time, AWS focuses on increasing the availability and reducing the cost of generative AI to improve the viability of use cases in the telco domain. With a broad choice of foundational and large language models available through proven AWS services such as Amazon SageMaker Jumpstart and Amazon Bedrock, developers of telecom operators can access different LLMs easily and securely at the commercial price points offered by cloud consumption models. AWS works with leading partners to provide GPU resources to run both training and inference effectively and significantly reduce costs through its own silicon innovation with Trainium and Inferentia.
Paths to Leveraging Generative AI
Operators will consume AI through productivity applications and enterprise software, but the greatest potential value comes from combining private telecom data with generative AI. To achieve this, operators can pursue multiple approaches:
- Pre-Trained Models: Incorporating private data in prompt contexts for pre-trained foundation models is the quickest way to achieve results without extensive customization.
- Fine-tuning Models: Adapting pre-trained models to telecom use cases by fine-tuning with private data offers a balance between customization and speed, potentially yielding higher accuracy and performance.
- Custom Model Training: Training custom foundation models from scratch using large telecom datasets can provide proprietary capabilities and maximize competitive advantage, but it requires significant time and investment.
We recommend starting with pre-trained models and fine-tuning them before embarking on building models. Operators should evolve their approach based on use case requirements, team capabilities, and the maturity of their data modernization efforts.
Beyond cloud services, telecom operators must recognize the importance of upskilling and retraining employees in cloud and AI/ML. AWS has established free training programs in cloud skills for 29 million individuals. Our AWS Generative AI Innovation Center, staffed by AI/ML scientists, solution architects, and other experts, works closely with developers and customers to facilitate AI/ML adoption and enable exciting innovation journeys in the telecom industry.
