WeaversBRAIN Unveils Action-Driving AI 'AgentMAX,' Targets Enterprise Market Beyond Education

WeaversBRAIN unveiled its AI agent platform, "AgentMAX." On the 3rd, WeaversBRAIN held a press briefing for the "2027 Tech Day" at the El Tower in Yangjae, Seoul, and announced the AI agent development platform "AgentMAX" and its business expansion strategy. AgentMAX is a platform for developing specialized AI agents by sector.

WeaversBRAIN introduced AgentMAX by leveraging its voice AI technology and user behavior analysis capabilities built on its language education services. AgentMAX uses accumulated learning and customer counseling data, and it is equipped with functions that guide user behavior.

WeaversBRAIN plans to expand AgentMAX beyond learning support into product sales and customer counseling. Its target market is enterprise AI solutions. Services based on AgentMAX include learning support tool "MaxTutor" and purchase decision support tool "MaxSales."

The two services work by having AI initiate the conversation first and suggest the next action when learning delays or hesitation over a purchase occurs. The focus is not on handling tasks on a user's behalf, but on helping users continue learning and make purchase decisions themselves.

WeaversBRAIN presented this structure as its own concept of AI agents. Jo Se-won, CEO of WeaversBRAIN, explained that existing AI agents have handled people's work in their place. He then said that studying, making purchase decisions, and exercising are things people must do themselves. He added that guiding such behavior is what WeaversBRAIN defines as an AI agent.

WeaversBRAIN also disclosed the actual response flow in a purchase context through a demonstration of an e-bike sales website. Kim Min-sam, Chief AI Officer (CAIO) of WeaversBRAIN, demonstrated MaxSales' customer response process on the sales website for the affiliate's e-bike brand, VIAGGIO.

During the demonstration, the presenter hesitated while checking the price. At that point, the AI initiated the conversation first. It then continued the interaction by asking about the intended use, such as commuting or leisure.

The AI then displayed estimated travel times by distance on a map and recommended products with suitable battery capacities accordingly. However, the final purchase decision was left to the user.

During the response process, the system handled questions and answers about charging time and uphill riding performance. When price concerns arose, it responded with an explanation converted into monthly payments and information on applicable promotions.

When hesitation over a purchase persisted, it suggested applying for counseling, and customer information handling was reflected in the input form. In this way, the system moved the user to the next step through a counseling application and information entry when purchase hesitation occurred.

Kim, the CAIO, said the agent generates real-time screens and directly operates web pages to support customer decisions. Based on customer voice recognition, MaxSales determines its response method according to conversation context and provides the necessary information by voice while also controlling the web page. This function uses more than 10 years of accumulated sales and counseling data.

However, consent for collecting personal information was left to the user's direct choice, and the final execution of the purchase button was also classified as something the user must do directly. The company said AI can support purchase decisions, but the criterion for whether an actual purchase is made is the user's own judgment and resolve.

The company emphasized the ongoing accumulation and updating of its existing sales data as part of its efforts to use it. CEO Jo said it is using sales data accumulated over more than 10 years for AI training. He said the company is also securing new counseling data and applying the secured data to advance the technology. Through this, he said, it has secured data-based technology.

As examples of products that were either nonexistent in the past or not widely popularized, Jo cited e-bikes and AI glasses. He said that in those product categories, AI analyzes product features and also analyzes competing products. He added that, based on those results, the company is helping customers choose the most suitable counseling approach.

The company said the scope of the demonstration disclosure was customer response functions. It also explained that it is developing product information research technology and technology that selects the appropriate method among various conversation paths. The current application of MaxSales is the sales website for the e-bike brand VIAGGIO, it said.

However, it said the purchase conversion rate is undisclosed. Sales contribution is also undisclosed. The learning continuation rate is 63%, it said. AI costs have been cut by 91%, it said.

WeaversBRAIN is pursuing an AI agent business based on its experience operating language education services. The company has motivation technology designed to prevent learners from procrastinating or dropping out, and it determined that the technology could be applied to other industries. CEO Jo said that people tend to hesitate, procrastinate, and forget, and defined a true personal assistant AI agent as one that helps users restart at the moment they stop.

Before the presentation, a phone lesson demonstration of the AI English tutor "Leo" was held. In the demonstration, the learner tried to postpone the class by saying they were in a meeting. In response, the AI suggested investing 30 seconds to practice one English sentence, and after the brief lesson, a follow-up call was scheduled for the evening.

WeaversBRAIN said the learning continuation rate for subscribers to "SpeakingMax" after 3 months is 63%. It also said it has 1.23 million cumulative members and has secured 680 million utterance data points. It has also accumulated more than 10 years of counseling and sales data.

The company is using its cumulative members, utterance data, and counseling and sales data for AI agent development. WeaversBRAIN is developing AI agents based on its operational experience helping learners continue studying in language services and on data accumulated over a long period of time.

The company said it has independently secured and internalized conversational AI technologies such as voice activity detection (VAD), speech-to-text (STT), and text-to-speech (TTS), and that since 2023 it has operated a separate AI research organization, AI Lab, to advance related technologies. Through this, it said, it cut the cost of using external AI technologies by 91% by using its own technologies.

Against this backdrop, the core of AgentMAX is its behavior design engine, EDRA. EDRA is composed of four stages: Empathy, Diagnosis, Reframing, and Action, and it identifies the reason for a user's hesitation and suggests actionable goals.

WeaversBRAIN explained that while existing large language models (LLMs) are centered on responding to user commands, AgentMAX is centered on prompting behavior to achieve long-term goals. To this end, it fine-tuned related models, including speech recognition, language models, and voice responses, using education and sales data.

WeaversBRAIN said data accumulated in education settings and sales processes is the basis for expanding its AI solutions. CEO Jo cited the learning and counseling data accumulated over a long period as the background for expanding the company's existing edtech business into AI solutions, and said it judged that the data secured from education services and sales sites could be used in other industries.

In line with this trend, the company said it is supporting learners through preemptive phone-based learning using "MaxTutor," developed on the basis of AgentMAX, in the education sector. The company said the cumulative number of organizations that have adopted "MaxTutor" stands at about 300 companies and institutions.

WeaversBRAIN signed a memorandum of understanding (MOU) with NE Neungyule on the 2nd for the supply of MaxTutor and the joint development of customized AI services. It is also pursuing the development of customized language learning AI with Hankuk University of Foreign Studies and plans to expand the scope of services to other subjects such as math in the future.

In the non-education sector, WeaversBRAIN signed a sales partnership with Even Realities. It plans to use MaxSales for domestic AI glasses sales.

This partnership is the first case of using AI agents outside the education sector. WeaversBRAIN outlined a plan to expand the technology applied to existing e-bike sales into the AI glasses market.

CEO Jo said the company plans to provide tutor agents to other universities and education companies. He also signaled his intention to expand the supply of AI solutions to external educational institutions. In line with this, WeaversBRAIN plans to supply technology proven in its own services to outside companies and expand its enterprise AI agent solution business.

WeaversBRAIN and SK Telecom (SKT) are jointly developing an AI agent for the education sector within the "Everyone's AI" project. CEO Jo said related cooperation is under way and explained that the company plans to continue its relationship with SK Telecom (SKT). WeaversBRAIN is also participating in the K-AI Alliance led by the SK Group.

WeaversBRAIN plans to conduct its business by combining AI agents with partner products and services rather than developing devices on its own. Based on this, it is reviewing ways to expand into vehicle infotainment and home Internet of Things (IoT) applications.

WeaversBRAIN's 2025 revenue came to KRW 41.5 billion. That was up 23.5% from the previous year. Operating profit was KRW 3.27 billion, up 36.3% from the previous year. WeaversBRAIN posted profits for 5 consecutive years.

The company is expanding from language education into enterprise AI solutions, and the key indicators for gauging its future business competitiveness will be securing external clients and actual purchase conversion performance.

Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=409190

References

• https://www.epnc.co.kr/news/articleView.html?idxno=409190