DeepBrain AI Develops 'Customer Service AI Avatar' for SMEs... Pilot Testing With 4 Companies Next Year

DeepBrain AI announced on the 28th that it has begun developing a 'customer service AI avatar' for small and medium-sized enterprises (SMEs). DeepBrain AI has already applied its conversational AI avatar technology to banks and educational institutions, and this time it is expanding the scope to customer service operations at SMEs.

The service is aimed at SMEs that face surges in customer inquiries or have difficulty securing staff for always-on customer support. It will operate by supporting AI avatars in customer-facing tasks by industry, with example use cases including guiding visitors at furniture showrooms, handling exam-related inquiries at educational institutions, customer support for food companies, and technical support at industrial sites.

DeepBrain AI said it adopted a design principle of differentiating the avatar's role and response style by task characteristics. Accordingly, it plans to develop the service in a way that assigns different roles to suit customer-facing tasks that vary by industry.

DeepBrain AI plans to complete service requirement definition and screen and feature design by the end of this year. Next year, it plans to establish real-time video transmission capabilities and web, mobile, and kiosk usage environments, and to conduct pilot testing with 4 companies.

This structure uses PosiCube's agentic AI contact center (AICC) core model in the process of analyzing customer questions and generating answers. PosiCube AI determines responses to each inquiry and the work required. DeepBrain AI's avatar engine then delivers those judgments in real time through video and voice.

The AI avatar converts generated responses into speech. It also renders mouth movements, facial expressions, and gestures matched to the utterance. It is designed to begin speaking immediately after an answer is generated, and the design focus is on reducing user waiting time.

For video and voice transmission, the company will apply web-based real-time communication technology, webRTC. The service will be available through web browsers, iOS, Android applications, and in-store kiosks. It plans to provide the same AI avatar service regardless of the operating environment.

This structure goes beyond responses and also connects to post-consultation tasks. Consultation records are automatically entered into the CRM system. If actual tasks such as ordering, refunds, or ticket creation are needed, the follow-up process will be carried out after manager approval.

The service will be pursued by setting the avatar's role and tone of voice differently to match industry characteristics. In furniture and food, it will reflect the role and tone of a customer service representative; in education, those of an instructor; and in energy, those of a facility expert. If a user is recognized as a foreign customer, the company will also develop a function to switch to a multilingual avatar.

Pilot testing will be pursued next year. 4 SMEs in the furniture, education, food, and energy sectors will participate, and the service will be applied to each company's actual customer-facing operations. The evaluation items are response quality, latency, and follow-up task linkage functions.

Vivipi is a furniture company that will apply the AI avatar to its showroom kiosks. The AI avatar will provide guidance on product specifications, quotations, and AS reception. Edupotec, an education company, will handle inquiries that surge during specific periods with an instructor-style AI avatar. The inquiry types are exam schedules, refunds, and grades, and objections that require separate judgment will be connected to a staff member.

PoviHours is a food company that will use inquiries from multiple channels, including its own online store and open markets, in an integrated way. PoviHours will use this to guide users in searching for country-of-origin and allergy information. It's Solution, an energy management company, will use an expert-style AI avatar to explain response procedures when signs of equipment anomalies occur. For urgent failures, It's Solution will pilot the service in a way that connects users to a real expert.

DeepBrain AI plans to expand its supply scope based on the 4-company pilot results. The expansion target is other SMEs in the same industries. Companies without dedicated IT staff are also included. The company is developing automation features for the initial deployment process with the goal of enabling service setup within a day once the necessary data is registered.

This concept contrasts with existing large-scale AICC systems, which require customized deployment for each company. DeepBrain AI is planning a method that eases the initial adoption burden for SMEs compared with existing large-scale AICC systems. In the operational phase, it will apply a role-splitting approach between humans and AI, in which AI handles repetitive customer service tasks and staff are connected when judgment or approval is needed.

DeepBrain AI is supplying conversational AI avatars to major financial institutions and cyber universities. Last year, the Korea Laboratory Accreditation Scheme (KOLAS) conducted a comparative evaluation of products from 5 overseas companies and 3 domestic companies. The comparison covered video quality, lip-sync accuracy, and pronunciation error rates. The company said its products received top evaluations in major categories.

Seo Jeong-hyeon, a director at DeepBrain AI, said the company redesigned the conversational AI avatars it has supplied to financial institutions and cyber universities for use at SMEs that lack sufficient customer service staff. He added that the company will pursue pilot testing next year and verify response quality and response delays against on-site standards during the pilot to complete a form that can be supplied in practice.

Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407433

References

• https://www.epnc.co.kr/news/articleView.html?idxno=407433