Contributed Article
by Dr Ishwar Parulkar, Chief Technology Officer for the Telco Industry at AWS
Generative artificial intelligence (AI), a type of AI that is capable of producing new content such as conversations, stories, images, videos, music, and code, continues to make headlines on how it will transform industries – telco included.
Anticipated growth is substantial. In fact, our own survey of telco leaders, facilitated with Altman Solon, found that adoption of generative AI use cases will grow from the current 19% adoption to 48% within the next two years.
While there are many use cases for how telcos can embrace generative AI, we see adoptions occurring in three waves.
Telco use cases across three waves of adoption
The first wave will take advantage of foundation models and capabilities that exist today, largely in customer experience (CX). Telcos already leverage AI to enhance interactions and resolution speed, and according to our survey, almost all (92%) respondents selected customer service and chatbots as a high likelihood to implement. In fact, among those respondents, 63% said the deployment was already in production. Embracing generative AI can further this progress with interactive voice response and real-time call analysis to provide prompts and resources for agents to help resolve customer inquiries. Customer service agents will still play a key role in the process, but generative AI can improve every customer interaction.
The second wave will comprise fine-tuning foundation models for telco purposes using proprietary data. One early example of this is the work Snowflake and DigitalRoute recently showcased to combine data from billing support systems (BSS) and operational support systems (OSS). Using Llama 2, an open-source foundation model fathered by Meta, and trained using Amazon SageMaker, this solution can help telcos more easily pinpoint and resolve network performance issues impacting key customers.
This can also be applied to challenges such as revenue leakage or optimising profits. At TM Forum’s Digital Transformation World, Salesforce showcased a new demo that uses Amazon Bedrock and Amazon Sagemaker Jumpstart to fine-tune models and combine data for use cases including revenue assurance, employee dispatching, and empowering customer service agents with meaningful real-time insights.
The third and final wave will be focused on creating new industry specific foundation models trained on telco specific data, for example standards specifications and data from the network and its operations. While two-thirds of telcos (65%) anticipate training off-the-shelf models to meet their needs, a cohort of 15% indicated a desire to build foundation models in-house.
We see opportunities for independent AI software vendors and early telco adopters of this technology to work together to use network data to build wholly new foundation models that can address network function software design, network design and configuration and network failure resolution related use cases. We’re already seeing some early movers in this space. For example, SK Telecom, Deutsche Telekom, e& and Singtel announced a Global Telco AI Alliance and collaboration with Anthropic to develop a new global telco-focused large language model. Not only will these efforts benefit the telcos and the industry, it also creates a new potential line of revenue for telcos to monetise their assets.
Generative AI begins with a data strategy
Regardless of which wave of use case telcos pursue, the most important piece is a solid foundational data strategy. Generative AI is only as good as the data it uses and the platform it is built on. Our survey found that organisations ranking in the top 30% for data proficiency are outpacing their peers in using generative AI.
Yet equally important is protecting that data. For some generative AI use cases, telcos need to customise existing large language models (LLMs) using company proprietary data. In using these publicly available LLMs, there is concern that proprietary company data could be embedded into the public model itself, creating intellectual property risk. Two-thirds (61%) of surveyed telcos indicated concerns around data security, privacy, and governance. Business and IT leaders should therefore work hand-in-hand with their security, compliance, and legal teams to identify and mitigate these risks, ensuring the secure and responsible deployment of generative AI. Moreover, businesses should carefully plan for compliance with regulations and consider the ownership of the data used.
Before commercially deploying any kind of AI application, it is key that businesses consider their existing data organisation and data platform strategy, and assess the expected return on investment. Certain applications will deliver greater impact depending on the available data. That said, we firmly believe that AI represents the most profoundly transformative technology of our era, with generative AI opening doors to incredible new opportunities that every business in the UK can and should consider tapping into.
Want to keep up to date with the latest developments in the world of telecoms? Subscribe to receive Total Telecom’s daily newsletter here
Also in the news:
Stonepeak buys minority stake in Cellnex Nordics
Telefonica Germany partners with Skylo for satellite-supported IoT
Sky Mobile network outages linked to removal of Huawei equipment