WeaversBRAIN

Beyond conversation,
AI that drives action

Tutor
Tutor. Sales. Care.

Even with GPT,
you’re still not
practicing English!

agent max

Beyond conversation,
AI agents that drive action

They reach out first, see it through,
and prove it with results.

  • Messages go unopened
  • Drop off before deciding
  • Too many people to respond to
  • Can't verify follow-through

Agent Incubator

AgentMAX is an incubator
for agents that drive human action.

We combine behavioral data and field know-how with our own conversational AI.
Our agents speak in the right voice for the moment, show the visuals that help, and get people to decide and follow through.

Tacit knowledge

Know-how unique to WeaversBRAIN

Teaching
How star instructors get learners unstuck
Sales
How top sellers answer each type of pushback
Coaching
How experts design habits

Data

Specialized data big tech doesn't have

1.23M
Total members
680M
Spoken sentences collected
14×
14× higher long-term retention than learners who don’t
Calls
Diagnosis of successful cases
agent max
  • VADVoice detection
  • STTSpeech recognition
  • LLMDialogue reasoning
  • TTSSpeech synthesis
  • Korean code-switching

    Handles mixed Korean and English

  • Turn-taking

    Knows when you're done and takes its turn

  • Full duplex

    Responds instantly, even when interrupted

  • Target speaker extraction (TSE)

    Hears only the user, even when many people talk

+ Visual aids

Shows the screens and materials the conversation calls for.

+ EDRA behavior design engine

Empathy → Diagnosis → Reframe → Goal breakdown

Swap the role, knowledge, and target behavior, and a new agent is born.

Agents that drive action

  • MAX Tutor
  • MAX Sales
  • MAX Care

+ Next agents

  • MotivationLearning · Beauty · Fitness
  • SalesEducation · Finance · Commerce
  • PublicAdministration · Healthcare · Welfare
  1. Measure Action Rate
  2. Feed results back as data
  3. Smarter interventions

Action Rate: the share of users who actually act after a conversation — AgentMAX's core performance metric.

How it works

From signal to action

The agent reads the situation and moves first,
even before the user asks.

  1. STEP 1Detect signals

    Catches hesitation, stalls, and drop-off signs in real time

  2. STEP 2Pick the moment

    Chooses when the user is most likely to respond, based on their routine

  3. STEP 3Converse

    Guides the user to act with a set persuasion sequence (EDRA)

  4. STEP 4Confirm follow-through

    Checks whether the user acted and plans the next move

  5. RESULTAction & achievement

    Measures the Action Rate and refines the next intervention

Difference

How it differs from information chatbots

Information chatbotAgentMAX
How it worksAnswers questions passivelyReads the situation and steps in proactively
DataMostly public documents and general knowledgeBehavioral data recording completion and conversion
PurposeSummaries and Q&APersistence, decisions, achievement
When it actsOnly when the user speaks firstThe moment drop-off or hesitation is detected
Unit of actionLong text explanationsTasks you can do in under a minute
ContextLost between sessionsCarries over across sessions and channels

Why us

Anyone can use an AI model,
but the real edge begins with data on whether people actually followed through.

Persistence data

We've tracked who kept learning and who stopped
across years of running language services.

  • What we said to win people back right before they quit
  • When re-engaging drew the strongest response

Decision data

We've recorded why sales succeeded or failed
across 15 years of consultation teams.

  • Responses that converted each type of refusal
  • When and how an offer brought the decision forward
1.23M
Total members
300+
Companies & institutions
680M
Spoken sentences collected

WeaversBRAIN brings together behavioral data,
domain experts, and AI research to build

agents that drive action.

agent max

An AI agent that speaks first and drives action

You've put it off 3 times this week.
Is the timing off?

Looking for a commuter e-bike?
A foldable, long-range
model fits best.

It's pretty chilly today.
Are you doing okay?

Calling at the time we set.
Got 10 minutes now?

Time for your lunchtime meds.
Do you have them nearby?

Use Cases

Teaches like a star instructor,
sells like a top salesperson

AgentMAX reads people
and drives action in under a minute.

  1. Empathy
  2. Diagnosis
  3. Reframe
  4. Action
  • Star instructor mode

    MAX Tutor

    Calls first at the set time
    and gets the day's lesson done.

  • Top salesperson mode

    MAX Sales

    Steps in when buyers hesitate
    and helps them decide on their terms.

  • Head coach mode

    MAX Care

    Reads diet and daily patterns
    and changes the very next choice.

MAX Tutor

Star instructor mode

These conversations are recreated to show the tone of the actual service.

Trust & Control

Control stays with the user

The more an agent drives action, the more users
should decide how far and how hard it steps in.

  • Data isolation

    Personal data and business information are kept separate, and are never used to train external public models without explicit consent.

  • Set the scope

    Users decide when they can be contacted, how strong notifications are, and what data the agent may refer to — and can change these at any time.

  • Approval for what matters

    Routine guidance runs on its own, but high-stakes actions like payments and contracts require the user's approval.

Partners

Partnerships and recognition

  • SK Telecom

    SK Telecom

    AI agent business MOU,
    'AI for Everyone' collaboration

  • SK K-AI Alliance

    SK K-AI Alliance

    46 AI companies
    from Korea and abroad

  • Ministry of Science and ICT

    Ministry of Science and ICT

    Selected for the
    high-performance
    GPU support program

  • Even Realities

    Even Realities

    AI glasses sales agreement

  • NE Neungyule

    NE Neungyule

    Technology supply agreement

  • Flexcil

    Flexcil

    Technology supply agreement

  • Hankuk University of Foreign Studies

    Hankuk University of Foreign Studies

    Industry-academia partnership on multilingual AI tutors

Agents configured for your environment

We review your business goals and data environment, then design the scope and approach together.

Get in touch

Service

Services powered by our agents

max ai

MAX AI

Motivation that never stops — No.1 AI English

Visit website
SpeakingMAX

SpeakingMAX

Real-world English filmed on location — No.1 in customer satisfaction

Visit website
VIAGGIO

VIAGGIO

Buying advice that reaches out first — premium e-bikes

Visit website

About

WeaversBRAIN

We are an AI company that builds conversational AI agents that drive human action, for our own services and for companies and institutions.

WeaversBRAIN
Company overview
CompanyWeaversBRAIN Co., Ltd.
FoundedJuly 2008
CEOSewon Cho, Yongkook Lee (Co-CEOs)
Address9F Mario Tower, 28 Digital-ro 30-gil, Guro-gu, Seoul, Korea
BusinessDevelopment and supply of conversational AI agents; AI language learning services
ServicesMAX AI, SpeakingMAX

History

Our story

We designed our own dialogue reasoning technology and built a voice AI pipeline.

  1. 2026

    Built a conversational AI pipeline

    • Secured core conversational voice AI technology by developing bilingual text-to-speech (TTS) in-house and deploying it in production
    • Launched SpeakingMAX Chinese and MAX AI Spanish, Vietnamese, and Korean courses
    • Signed an AI agent MOU with SK Telecom and a multilingual AI tutor MOU with Hankuk University of Foreign Studies
    • Joined the SK K-AI Alliance; selected for the high-performance GPU support program
  2. 2025

    Expanded the AI service lineup

    • Launched Care Call, where the AI tutor calls learners first
    • Launched SpeakingMAX Japanese, a MAX AI OPIc course, and domain-specific services for businesses
    • Bronze at the AES Global Awards; Minister of SMEs and Startups Award at Inno-Biz Day
  3. 2024

    Became an AI company

    • Renamed the company WeaversBRAIN
    • Launched MAX AI, an AI tutor built on our in-house dialogue reasoning technology

News

News from WeaversBRAIN

Loading news.

Contact

AgentMAX inquiries

Sales agent

For companies and institutions considering agent adoption, voice AI engine integration, or joint projects.

Get in touch

Tutor agent

Learn how to bring employee language training and domain-specific courses to your organization.

Get in touch

Join us in building
AI that drives action

Open positions are listed on
the Weavers Group careers page.

Go to careers page
기업소식

위버스브레인, 과기정통부 '첨단 GPU 활용 지원사업' 선정

위버스브레인, 과기정통부 '첨단 GPU 활용 지원사업' 선정…블랙웰 B200 인프라 확보

▷ 6억 8,000만 건 발화 데이터를 더 자연스러운 대화로…B200으로 자체 음성 엔진·LLM 동시 고도화
▷ 기존 대비 한층 큰 연산 자원 확보…대화형 AI 기반 신사업 확장 추진

• 과기정통부·NIPA 주관 '첨단 GPU 활용 지원사업' 최종 선정…블랙웰(B200) 인프라 확보
• 6억 8,000만 건 누적 발화 데이터 기반 대화형 AI 모델 학습·고도화에 투입
• 자체 음성 엔진과 파인튜닝·블렌딩 대규모 언어 모델 동시 고도화

AI 교육 기업 위버스브레인(대표 조세원·이용국) 은 과학기술정보통신부와 정보통신산업진흥원(NIPA)이 주관하는 '첨단 GPU 활용 지원사업'에 최종 선정됐다고 밝혔다.

이 사업은 인공지능(AI) 연구·개발에 필요한 고성능 컴퓨팅 자원을 국내 중소·벤처기업에 지원하는 정부 사업이다. 위버스브레인은 이번 선정으로 블랙웰(B200) 기반 GPU 자원을 배정받았다.

위버스브레인은 확보한 컴퓨팅 자원을 대화형 AI 모델 학습과 고도화에 투입한다. 자체 음성 모듈을 고도화하는 한편, 파인튜닝과 블렌딩으로 만든 교육에 최적화 된 대규모 언어 모델(LLM)도 함께 고도화한다.

이 같은 학습은 위버스브레인이 영어 학습 서비스 스피킹맥스, 맥스AI 등을 운영하며 축적해 온 6억 8,000만 건 누적 발화 데이터를 기반으로 한다. 위버스브레인은 이 데이터를 활용해 더 자연스러우며 대화 상황에 맞춰 톤을 조절할 수 있는 대화형 AI로 고도화한다는 계획이다.

위버스브레인은 음성 감지(VAD)·음성 인식(STT)·음성 합성(TTS)에 이르는 대화형 AI 파이프라인 전 구간을 자체 기술로 내재화하고 있다. 지난 5월 음성 합성(TTS) 모델을 자체 내재화했고, SK그룹의 'K-AI 얼라이언스'에 합류했다.

조세원 위버스브레인 대표는 "기존에 활용하던 인프라보다 한층 큰 연산 자원을 확보하게 됐다"며 "누적 발화 데이터를 바탕으로 더 자연스럽고 톤 조절이 가능한 대화형 AI를 고도화 하는데 전력을 다할 것"이라고 말했다.

관련기사 ▶▶ 머니투데이|"더 자연스런 대화 AI 개발"…위버스브레인, GPU 지원사업 선정 (2026.06.16)

※ 기사 제목 클릭 시 원문으로 이동합니다.
WeaversBRAIN

Your browser needs an update

This site doesn't display correctly
in the browser you're using.
Please use an up-to-date browser.

Get ChromeGet Edge

On iPhone or iPad, update iOS in
Settings › General › Software Update.