An AI framework capable of combining the power of multiple AI models
The AI-driven EBITDA mastery: Revolutionizing customer journeys Catalyst is developing an AI-powered framework that uses a federated data model which can be applied to multiple use cases across the industry
An AI framework capable of combining the power of multiple AI models
Commercial context
A crucial aspect of CSPs' modernization strategies involves investing in new technologies that enhance value, streamline operations, and aid in developing and retaining customers. Across the industry, AI is widely recognized as a powerful solution to address these challenges. However, implementing AI to boost operational efficiency and transform the costliest aspects of customer and service journeys into profitable, competitive differentiators is a complex endeavor.
The fragmented and siloed nature of CSPs' data often hampers their ability to use AI effectively - a systemic issue that needs improvement. The solution lies in developing an AI-powered framework that employs a federated data model and multiple AI models that can enhance response value across various domains and customer lifecycle stages without requiring extensive data migration.
The solution
The AI-driven EBITDA mastery: Revolutionizing customer journeys Catalyst project, which won a 2024 Catalyst Award in the Best Moonshot Catalyst - AI Challenge category, is designed to produce such a framework. Based around an open, composable AI architecture capable of accessing diverse AI models and a federated data model, the framework centralizes some data while accessing other data via OpenAPI calls, enabling CSPs to fuel AI initiatives efficiently. The framework is underpinned by various TM Forum assets: the TMF678 Customer Bill Management API is crucial for the information flow in retrieval augmented generative AI, GB998 ODA Concepts & Principles align the functional architecture, and IG1171 ODA Component Definition v5.0.0 further refine it. The project demonstrates its effectiveness through two key use cases, both designed to hyper-personalize customer journeys and significantly boost efficiency and customer experience.
The first application, sentiment-aligned AI billing inquiry management, tackles the high volume of billing-related inquiries, which constitute 50% of customer care calls and drive significant operational expenditures. By employing AI to manage these inquiries, CSPs can reduce the total cost of agents, improve first-call resolution rates, decrease average call handling times, and proactively address issues before they lead to customer calls. This approach not only enhances customer satisfaction but also boosts sales and upsell conversion rates, reduces churn, and improves employee experience by lowering turnover rates. Moreover, it aims to cut down on credits and discounts resulting from disputed charges.
The second use case focuses on field service management, addressing the high OPEX costs associated with field service truck rolls, which can account for 17% to 30% of a CSP's operational expenses. By using historical diagnosis and real-time information through generative AI (GenAI), the framework aims to accelerate work order completion, guide field engineers through solutions, and reduce the need for helpdesk intervention. This approach ensures adherence to service level agreements, mitigates fines, and effectively communicates remediation plans to customers.
Wider application and value
The framework is showing all the signs of playing a transformative impact for the wider telecoms industry. Although still in its early phases, the project team estimate that the framework can reduce OPEX by 10.7% to 20.7%, which translates into an EBITDA growth of 31% to 57%, a significant leap in an industry where EBITDA growth has stagnated at around 0.6% on average. The first use case alone could deliver a 3% to 9% increase in EBITDA by reducing customer service OPEX by 15% to 26%, while the second use case contributes approximately a 2.6% increase in EBITDA through a 7% reduction in network operations OPEX.
As Eugene Yeo, Deputy CEO of Converge ICT explained “this composable and open GenAI-based architecture is scalable to numerous other use cases across sales, marketing, customer support, and network operations. By using additional federated datasets via dynamic function calling, the AI system learns which TM Forum Open APIs to call, delivering significant efficiencies and further reducing OPEX. As CSPs continue to struggle with high capital investment intensity and increasing operational expenses, this Catalyst initiative provides a compelling solution to improve profitability, attract equity investment, and provide the cash fluidity needed for future investments. Through the demonstrated efficiencies in handling time and operating expenses, the project underscores the transformative potential of AI in revitalizing the telecom sector's financial health and operational performance.”