SKT Successfully Operates AI Customer Center Beta Service... Full Implementation from Next Year

Operating AI Customer Center Beta Service Since Mid-Last Month
Applying Proprietary Telco LLM·LMM... Developing RAG as Well
Automating Post-Consultation Tasks Starting December

SK Telecom announced on the 18th that it has successfully operated a beta service for about a month by applying its self-developed artificial intelligence (AI) consultation support system to its customer center.

The self-developed AI technology incorporates telecommunications-specialized large language model (Telco LLM) and large multimodal model (LMM) technologies.

SKT has been developing the Telco LLM in collaboration with major overseas LLM companies such as Anthropic. Over the past year, it restructured its knowledge information system to be suitable for AI utilization and conducted model tuning using telecommunications-related data.

Additionally, dozens of customer center consultation experts participated in creating training data to improve the performance of the Telco LLM. Subsequently, through year-round testing and feedback-based reinforcement learning, they developed telecommunications service-dedicated LLM and LMM models.

Furthermore, SKT developed and applied retrieval-augmented generation (RAG) technology to enable the LLM to generate clear answers by searching accurate information and enhancing answer quality. Not only text but also images were learned, such as telecommunications-related documents, to develop a telecommunications-dedicated LMM capable of understanding images.

Starting next year, SKT plans to fully apply AI technology to all customer consultation tasks. The Telco LLM-based AI knowledge search assistant service allows consultants to quickly find necessary information and guide users. Consultants can derive information by asking questions in natural language like 'Perplexity' or 'ChatGPT' without needing to search the knowledge information system.

Using AI transformation (AX) LMM, SKT will also build an AI document automatic processing system. The customer center handles various types of documents such as photos and scanned documents sent through multiple channels including text messages, emails, and faxes. AI is expected to reduce the time required to verify supporting documents by classifying and processing various forms and types of documents. The automatic classification system was introduced on the 23rd of last month, and technology to automatically read document contents will be applied in the future.

Next month, SKT will implement a post-consultation task automation system that automatically classifies, summarizes, and organizes customer consultation results using the Telco LLM. SKT stated, "The task of organizing consultation results used to take about 30 seconds on average, but with AI automatically processing it, consultants can quickly provide consultations to waiting customers without follow-up work after the consultation."

Hong Seung-tae, head of SKT’s Customer Value Innovation, said, "As SKT evolves into an AI customer center, we are not only focusing on improving work efficiency but also aiming to provide consultations optimized for each individual customer. Using the Telco LLM-based consultation support system as a stepping stone, SKT will continue to utilize various AI technologies in customer consultation tasks to provide the highest level of customer center consultation services in Korea."