Tenaga Nasional Berhad (TNB) transforms customer experience with AI on AWS

TNB's customer service model relied on a centralized contact center service, TNB Careline, where customers contacted live agents to report service issues, update personal information, or ask routine questions. With a growing customer base and increasing expectations for near real-time, 24/7 support, the utility sought to evolve its approach beyond traditional call-based interactions.

To improve service accessibility, TNB had previously introduced Ask Lisa in 2018, a keyword-driven FAQ tool. While Ask Lisa provided basic self-service functionality, it was not designed to interpret natural language or support more complex, conversational queries. With Malaysia being a multilingual country and customers communicating in English, Bahasa Malaysia, Mandarin, Tamil, and regional dialects, the utility recognized the need for more intelligent, multilingual engagement.

Under the leadership of President and CEO Datuk Ir. Megat Jalaluddin Megat Hassan, TNB identified customer engagement as a core part of its broader commitment to technology-led innovation. This direction laid the foundation for developing SARA, a virtual assistant designed to deliver consistent, AI-powered support and replace Ask Lisa. SARA offers 24/7 availability and is accessible across languages, regions, and service types. TNB built the solution using Amazon Bedrock, a fully managed service for building and scaling generative AI applications securely.

TNB worked with AWS and Uberfusion Sdn Bhd, an AWS Partner, to develop SARA, a generative AI assistant that delivers scalable, front-line support across the utility's digital channels. Designed with a friendly, personal and conversational flow, SARA simplifies customer interactions by avoiding rigid menu structures. Uberfusion worked alongside TNB to build and deploy SARA, helping to configure the model, integrate backend systems, and maintain the virtual assistant’s performance at scale.

The jointly developed solution now serves as the first point of contact for three primary use cases: answering FAQs, updating customer details, and reporting outages or service disruptions. SARA integrates with TNB's outage management system, TOMS, CRM platforms, and TNB Careline for live agent escalation.

Customers can report outages, streetlight faults (with phone verification), and manage their account information directly through the AI assistant in a secure, streamlined manner. When a query requires human support, SARA seamlessly routes customers to TNB Careline’s agents, thus maintaining service continuity. Within 10 months of SARA’s launch, this functionality has reduced reliance on live agents, resolved 15,000 queries monthly through AI self-service, and contributed to a 91 percent positive customer sentiment score.

Using Amazon Bedrock, TNB accessed foundation models such as Anthropic's Claude 3.5 to power contextual, human-like conversations that improved customer interaction quality across multiple languages. TNB also used Amazon Bedrock Knowledge Bases for retrieval augmented generation, which allowed Anthropic Claude to access domain-specific TNB data and generate precise information about TNB's services, policies, and procedures. TNB also integrated Amazon Bedrock Data Automation to transform diverse, unstructured content into actionable data for SARA's generative AI applications.

Finally, to create safeguards, TNB utilized Amazon Bedrock Guardrails to detect and mitigate harmful content in both user inputs and model responses, which in turn helps SARA maintain appropriate, secure interactions across all conversations.

TNB's success with SARA has positioned the utility as a national leader in AI-powered customer engagement. The solution has been recognized by multiple industry bodies, including the Software Guide Association of Malaysia and the GovInsider Public Sector Innovation Award. SARA has also been validated through Infocomm Media Development Authority of Singapore's AI Verify framework, demonstrating alignment with responsible AI standards.

"This transformation marks a broader shift in how TNB sees its role. With AWS, we're evolving beyond delivering services to building personalized, AI-driven experiences engaging Malaysians on their terms," said Datin Maryana Abdullah Maamuom, Project Director (TNB Virtual Agent Project), Tenaga Nasional Berhad.

Looking ahead, TNB will continue to invest in expanding SARA's functionality. Planned enhancements include voice integration to support customers who prefer spoken interactions. This initiative is designed to serve rural and elderly populations more effectively. The utility also intends to integrate SARA with its billing systems to provide personalized account access and implement seamless escalation to live agents for complex enquiries.