[비즈한국] LG Uplus032640 is expanding its AI services by integrating the 'ixi-O AI Assistant,' which provides search assistance during calls, into its proprietary call app, 'ixi-O.' Following the 'ixi 1.0' launch, which featured on-device AI for voice phishing prevention, the company once again emphasized a "first-of-its-kind" title. If a user says "Hey, ixi" during a call, the AI intervenes in the conversation to look up information. The other party on the line can also hear this voice guidance over the phone.
ixi-O is planned to be expanded into a vertical model in the future, offering meeting minutes and summarization features for specific regions and professional groups. As industry peers, such as SK Telecom017670, begin to initiate AI monetization—including plans for charging for 'A.' (Adot) in the first half of next year—LG Uplus stated, "We are ultimately preparing monetization models," but reserved public disclosure of specific timelines, adding, "Prioritizing customers feeling value in the service comes first."
ixi-O Evolves into an On-Device 'AI Assistant'
"It is no exaggeration to say that LG Uplus is the first in the world to offer a feature that allows users to perform searches without leaving a call." (Choi Yoon-ho, Head of LG Uplus AI Agent Promotion Group)
"ixi-O 2.0 has evolved one step further from the on-device AI technology implemented by the previous version. Now, users can search for information based on their conversation content using secure natural language processing. This is likely a first in Korea." (Lee Sang-yeop, CTO of LG Uplus)
LG Uplus is embarking on full-scale differentiation of its AI services by adding generative AI-based assistant features to ixi-O. The partner is Google. LG Uplus held a press conference in Jongno-gu, Seoul, on the 13th to unveil the ixi-O AI Assistant service, which was developed in collaboration with Google Cloud.
Launched last November, ixi-O is entering the '2.0' stage with the addition of the AI assistant, building on the '1.0' phase that featured real-time voice phishing detection, call recording, and summarization. The ixi-O AI Assistant is an on-device AI feature that understands the context of a conversation and provides immediate search results when invoked during a call. If a user calls ixi while talking to a colleague or friend to ask about the weekend weather or travel time from a subway station to a meeting spot, the AI voice finds the information and shares the results.
Models that convert speech to text (STT), infer intent and context (LLM), find an answer, and then convert it back to speech (TTS) can be inappropriate for phone calls where real-time response is critical. ixi-O employs 'Voice-to-Voice' technology, which processes the audio signal itself to generate answers without fully converting speech to text.
CTO Lee Sang-yeop explained, "We determined that the 8-second delay typical of existing LLM-based models was too long, as we need to replicate the patterns of natural conversation. v2v technology is an area where Google is ahead in terms of user perspective and search infrastructure, and we developed it in collaboration with Google using 'Gemini 2.5 Flash Live'."
v2v inference and search are handled by Google APIs. While it may take over 3 seconds depending on the network environment, the company explained that it is continuously improving latency by monitoring the entire process, including network optimization, app optimization, and server delays.
The LG Group possesses the 'EXAONE' AI model developed by the LG AI Research Institute. While the AI assistant feature uses external AI, the company plans to use EXAONE for on-device Small Language Models (sLM) to implement relationship inference and hyper-personalization, and ultimately build Agent Builder technology that expands autonomous actions.
Prioritizing Utility Amidst the Rush for Monetization
Recently, discussions on AI monetization have been expanding throughout the industry. The key is how to secure unique differentiators and create 'marketable AI' in both B2C and B2B sectors.
LG Uplus is conducting its AI B2B business primarily through its AICC (AI Contact Center) operations. The company plans to evolve its more flexible, LLM-based consultation solutions into a subscription model, moving away from the 'rule-based' systems that previously provided answers based on fixed scenarios. They are also pushing for overseas expansion to Southeast Asian telecommunications companies in the form of 'SaaS' (Software as a Service).
However, the company remained reserved regarding B2C monetization. Lee Jae-won, Head of the LG Uplus Consumer Division, stated, "We have no intention of charging separately for security and trust-related areas. However, we are contemplating how to create billing and monetization models for areas that further support customer convenience or productivity. Our priority is for customers to truly feel the value of the service and use it extensively."
Concerns regarding personal information collection and leaks are also anticipated, given that it is a call app that understands the context of private conversations and performs external (Google) searches. Regarding privacy, CTO Lee explained, "We implemented a 'Voice Secure Architecture' system that ensures conversation content is processed only on the user's device through on-device technology, and external inference is handled only during the search phase." The system sends only the prompted segments to the server to provide an answer, after which the data is deleted from the server.
The AI assistant feature has been available to Android and iOS users who requested it since the 12th. The goal is to secure 3 million users by next year. The company plans to continue improving the service in collaboration with Google, as the current model, focused on standard Korean, may struggle with heavy regional dialects. Choi Yoon-ho emphasized, "While other services are not yet linked, the technical readiness for partnerships is maturing. We will continue to expand our services alongside technological trends while monitoring customer feedback."