Matt Sanchez, Global Chief Data and AI Officer at Tecnotree and Vice Chairman at the Responsible AI Institute, on the ‘game changing’ potential Generative AI holds for telecommunications companies.
Generative AI is the most exciting breakthrough in the ever-evolving field of AI.
In telecommunications, Generative AI can be used to create content and insights that enhance customer experience, accelerate sales, improve operations and impact numerous key performance indicators (KPIs) like revenue, costs and customer satisfaction. It can bring about transformational changes and help telcos realize some of the promise of AI.
Generative AI Use Cases in Telecommunications
There are many business processes that stand to be enhanced with Generative AI.
Personalized customer experience and engagement, key for brand differentiation in an increasingly competitive marketplace, is one example.
Generative AI can be used to analyze customer data to generate personalized product recommendations, service adjustments, or promotional offers.
Customer call center, in store, and online and e-commerce interactions can all be transformed with personalized customer insights, interventions and content.
Generative AI models can also be used to drive customer service automation by improving AI-driven chatbots and virtual assistants to handle routine inquiries and tasks much better.
More automated responses to customer enquiries delivers time and financial savings. It can then help with the delivery of proactive service inquiry content, alerting customers with personalized messages about their service or billing needs, for example.
One of the most high-value uses of Generative AI in telecommunications however is in network optimization.
Complex networks that require continuous monitoring and optimization to ensure maximum uptime and quality of service can be enhanced by training AI models on historical network data to generate predictive models for network traffic, identify potential bottlenecks and suggest optimal network configurations.
Orchestration of multiple complex AI/ML models used for network planning and operations will combine with Generative AI models to understand network needs for capacity and performance.
Leveraging network performance predictions from supervised ML models, Generative AI can also help to figure out network maintenance plans.
Crucial to successful implementation of Generative AI are:
- Required Technologies
Large Language Models (LLMs) and Generative Adversarial Networks (GANs) are examples of Generative AI models used for creating new data by forming a complex and nuanced understanding of the training data.
GANs can be used to generate new customer journey scenarios or create ‘synthetic’ customers for testing service and product scenarios.
LLMs can derive insights from unstructured customer data, call center transcriptions, chatbot conversations and more – all with the goal of creating personalized customer service and engagement content. LLMs also can learn ‘in context’ through prompt engineering where a user provides relevant context in unstructured or structured form to the model as part of their query.
This ability to learn on the fly and even be further trained or fine-tuned to obey instructions makes LLMs and other pretrained models very flexible and powerful where precision and avoidance of incorrect data is as important as insights.
Embracing Cloud Technologies: Leveraging cloud platforms allows for the scalable storage and processing of large data volumes and flexible deployment of AI models.
Most Generative AI models take advantage of GPU (Graphical Processing Unit) compute to accelerate inference and training.
The state-of-the-art GPUs, like those from Nvidia, can be accessed quickly through cloud platforms thus providing an optimal deployment environment for Generative AI models. Many cloud providers are also offering Generative AI models as a service through an API ready for integration into customer applications.
Data Privacy Tools: Anonymization, data masking, and encryption tools are required to protect customer data. Ensuring that sensitive data is not leaked by Generative AI models is also a new frontier requiring special tools for security breach detection and monitoring.
- Key skills
Machine Learning and Data Science: The ability to process and analyze large datasets, including skills in statistical analysis, machine learning and proficiency in languages such as Python or R.
Understanding how to design, train and evaluate AI models, including knowledge of technologies such as GANs, NLP and recommendation systems.
Generative AI builds upon the general AI/ML practice where data science skills will remain critical for future successes.
Prompt Engineering: The formulation of input prompts that guide AI models such as Chat GPT to generate the desired outputs. A combination of prompt engineering and fine-tuning through techniques like RLHF (Reinforcement Learning through Human Feedback) are required to successfully deploy generative AI models in many cases.
Software Engineering: Developing robust, scalable and efficient software to implement and deploy AI models, requiring strong programming skills and knowledge of cloud platforms, data structures and algorithms.
Data Privacy and Ethics: Understanding legal and ethical considerations around data usage, ensuring regulation compliance and protecting user data.
Domain Knowledge: Telecom industry expertise, understanding its challenges, opportunities and customer expectations, is critical for solutions that enhance the customer experience.
Communication and Visualization: Communicating complex AI concepts to non-technical stakeholders and visualizing data and insights in a comprehensible manner.
By combining these technologies and skills, telcos can leverage Generative AI effectively to enhance and even accelerate their AI-powered transformations.
Risks and Challenges
Privacy and data security risk management is a priority.
Generative AI models require vast amounts of data for training, and this data sensitivity can pose a serious risk if not properly secured. The highest value use cases such as network optimization will require the use of proprietary data, which need data security and privacy capabilities – potentially novel given the extent of the use of Generative AI models.
Risks also revolve around the trust and governance of the insights produced by Generative AI-powered models and applications.
Telcos need to ensure their AI models are transparent, assessed for technical performance and trust and governable for a range of constituents.
Generative AI holds significant potential for telecommunications.
With the right approach, technologies, and skills Generative AI will be a game-changer for the telecommunications industry.
