![[한국IBM] Watson AI 소개 및 도입사례 (201904)](https://image.slidesharecdn.com/ibmwatsondigitalreinventionwithai201904-190625010614/85/IBM-Watson-AI-201904-32-320.jpg)
[한국IBM] Watson AI 소개 및 도입사례 (201904)
- 1. IBM Watson : Digital Reinvention with AI TaeGyu(Dave) Ahn IBM Data & AI “There’s no AI without IA – information architecture”
- 2. Watson Overview 2 Watson through cognitive enhances, scales, and accelerates human expertise IBM Confidential Understand Understands data–structured and unstructured, text-based or sensory–in context and meaning, at astonishing speeds and volumes. Learn Ingests and accumulates data and insight from every interaction continuously. Is trained, not programmed, by experts who enhance, scale and accelerate their expertise. Therefore, it gets better over time. Reason Has the ability to form hypotheses, make considered arguments and prioritize recommendations to help humans make better decisions. Interact Responds and communicates with people in a natural way that allows cognitive solutions to see, talk and hear. © 2019 IBM Corporation
- 3. Watson Overview 3 Watson @ Jeopardy 2011yr. & Watson Debater 2018yr. IBM Confidential © 2019 IBM Corporation
- 4. Watson Overview 4 Watson과 다른 인공지능 솔루션의 차별점 IBM Confiden+al • Customers develop and own insights with enterprise data and models • Customization capabilities for specific domains and private enterprise data • EU Model Clauses, ISO and PII* compliant; GDPR in 1H 2018 • Industry leading capabilities to learn from small data • Compare & Comply for understanding governing documents • Pre-trained smarts out of the box • • • & , A © 2019 IBM Corporation
- 5. Watson Overview 5 Watson APIs Portfolio IBM Confidential © 2019 IBM Corporation
- 6. Watson Overview 6 Watson Data Platform IBM Confiden+al IBM TOOLS Data Scientist Software Engineer AI Operations Build Run Manage 3RD PARTY IDE & FRAMEWORKS IBM AI RUNTIME 3RD PARTY RUNTIMES AI OpenScale Fairness Explainable AI Continuous Evolution Orchestration Model Health Accuracy Business User In app use your AI at scale With Trust and Transparency Watson Studio Watson Machine Learning InfuseIntoApps Infuse and Manage AI at scale © 2019 IBM Corporation
- 7. Watson Overview 7 IBM Confidential Watson is AI for professionals. Doctors & Pathologists Call Center Agent Lawyers Financial Analysist Field Engineers Store Associates HR Recruiter / Operator Digital Merchandizer Tax Professionals © 2019 IBM Corporation
- 8. Watson Overview 8 Watson through cognitive enhances, scales, and accelerates human expertise IBM Confiden+al Customer Engagement Process Automation Cognitive Insight • 챗봇 • 인공지능 콜센터 • 상품 및 서비스 추천 서비스 • 고객센터 오퍼레이터, 내부직원 Advisory (HR, IT Helpdesk, etc.) • 문서 분류 및 관리 • VOC(Voice of Customer) 분석 • 프로젝트 관리 지원 • 품질 및 정비 지원 • 지능형 탐색 지원 (마켓 센싱, HR, 세일즈, 보험 심사, 법률 서비스, 자산 관리 등) • 필드 엔지니어를 위한 지식 서비스 © 2019 IBM Corporation
- 9. Watson Business Solu.ons Watson Business Solution을 통해 기업고객이 신속하게 UseCase에 대한 확인을 하고 실제 Production을 구축 Watson APIs & Tooling Watson Business Solution Expertise, architecture, and pre-built assets Customer data + + 9 IBM Confidential © 2019 IBM Corporation
- 10. Watson Business Solu.ons 10 Introducing the Five Watson Business Solutions Voice of the Customer Allows clients to understand their customer’s sentiment and trends beyond what’s gathered from traditional surveys. Customer Care Provides the foundation for a new omni-channel engagement experience that allows clients to meet their customers in their channel of choice, while avoiding call center traffic. Expert Assist Helps business experts when they need more information to help solve a problem, answer a question or cross-sell a customer. Compliance Assist Extracts document elements and compares legal clauses to help professionals review and analyze contracts Take a picture using a mobile device and automatically identify what is in the image and specific characteristics (damages, features, etc). Identify objects and damages Take a picture and immediately extract relevant information and next steps about the identified object. Visual Discovery Quickly train custom models to help identify objects and specific conditions of the object. Quickly train custom models and… Visual Assist IBM Confidential © 2019 IBM Corporation
- 11. 현대카드 Buddy 롯데그룹 with WATSON B社 IT HELP DESK C社 IP/Patent Discovery 이마트 “띵구” § 현대카드의 카드혜택을 실시간으로 상담 § 콜센터 업무 Transformation으로 확대 § e-Commerce영역의 대화형 Digital Channel § Marketing업무 영역으로 확대 고려 § 맞춤형 Shopping Advisor 서비스 구현중 § Social / External data를 활용한 제품개발 § 글로벌 제조사의 IT HELP DESK 구축 § 전사 직원 HELP DESK로 확대 검토 § 제조회사의 신상품 개발을 위한 IP/Patent Discovery 서비스 (기존 Google 검색 대체) § 스타필드 고양 토이킹덤에서 운영하는 Watson기반 쇼핑도우미 로봇 § 쇼핑 Concierge BOT으로 확대 개발 A社 Commerce ChatBOT 국내 Watson 구현사례 및 Practices © 2019 IBM Corporation11
- 12. 1-800 Flowers.com® 은 Watson을 이용하여 선물 컨시어지 서비스를 제공하고 있습니다 12
- 13. 13 IBM Watson / © 2017 IBM Corporation Deakin 대학교의 학생 도움 서비스는 IBM Watson으로 연중무휴 서비스 되고 있습니다
- 14. IBM Watson은 노스페이스의 소비자가 보다 스마트한 방법으로 완벽한 점퍼를 찾을수 있도록 도와 줍니다 14
- 15. 월마트 노동조합은 노동자들을 돕기위해 IBM Watson을 도입했습니다 15
- 16. …적절한 시간에 제공되는 완벽한 의사결정 수익성 있는, 수 조원의 석유 시추 Watson technology culls through 30 years of Woodside’s documented knowledge. Engineers augment their expertise with a virtual advisor, “Lesson Learned.” Each outcome is added to collective knowledge, informing future decisions. G ROW K N OW L E D G E F RO M DATA E N H A N C E E X P E R T I S E L E A R N A N D A D A P T
- 18. 미국 A社 인공지능 기술을 IVR 에 적용하여 상담 요청 콜 수 감소 IBM Confidential § 1961년 설립된 미국의 의료 보험 회사로 의료 보험 소비자 서비스 판매 / 운영 § 미국 내에서 소비자 수는 1,150만 명, Rev. (’16) : $54B § Contact Center Call Volume : 170만 건/월, IVR completion Rate 18% IVR Completion Rate 를 높여서 상담 요청 호 줄이는 것이 주요 목표 프로젝트 개요 및 경과 결과 및 시사점 POC 현재 Phase1 완료 후Phase2 구축 프로젝트 수행 중 (2019년 1Q open 예정) • 음성인식(STT) à 대화(Conversa<on) à 음성변환(TTS) 로 Dialog를 통한 AI 기반 IVR 적용 • 의료 개인 정보 (PHI : Personal Health informa<on) 법령 및 가이드를 준수하는 아키텍처 설계 • IVR Comple<on Rate 56%로 개선효과 확인 • 업무 적용 영역 사전 정의 (전체 인입호의 70% 해당하는 문의 유형) 예) 보험 혜택, Membership, 가입 eligibility, Authoriza<on, Referral 등. • Dialogue 방식의 IVR 구현 : Dialog 체계를 사전 정의하여, 필요한 정보에 대해 User가 답변을 이어가는 방식 채택 • 적용 Watson API : STT, TTS, Watson Assistant 단계별 접근 사전 정의된 사용자 군 및 적용 영역에서 시작 대화 설계 lessons learned 구조화된 Dialogue 설계를 통해 대화의 흐름 통제 인공지능의 콜센터 적용 영역 확대 - STT & TTS 및 Watson Assistant 통합 구현 - 실제 대화하는 IVR 구현 © 2019 IBM Corporation18
- 19. 일본 P社 인공지능 기술을 IVR 에 적용하여 상담 요청 콜 수 감소 IBM Confiden+al IBM Voice Gateway After with VoicebotBefore Voicebot Solution Push Button IVR The customer was connected to the operator using the push button IVR system. Our client wanted to improve the connection tendency of the phone without increasing operators. Client had more pains and challenges : A. As XX Product became popular in Japan, the support calls increased. Customer's voice said that it was hard to connect on time. B. Call flow of Push Button IVR is complicated and XX Product User leaves halfway. C. Response quality depends on operator skill. Voicebot system understands the user's questions and returns a automatic response. This solution divides some questions that require operator response, pre-interview and handover the information to the operator. Voicebot solution solves: A. 9000 call/day can be handled without increasing operators. Callback reservation for 24 hours contributes to opportunity loss and satisfaction improvement. B. The user only has to talk as with the operator, and the withdrawal rate decreases. C. An accurate automatic answer by knowledge stored in Chatbot System. IVR Interactive Voice Response System 100 100100 30 70 long time waiting Quick response © 2019 IBM Corporation19
- 20. IBM Watson Conversation : 국내 금융권 도입 사례 20
- 21. IBM Watson Conversation : 국내 금융권 도입 사례 21
- 22. Watson Assistant (Conversation Service) Intent 정의 및 문장 학습 Entity 정의 및 등록 하나의 Intent에 대한 다양한 문장(표현)을 실제 문장에 가까운 문장으로 수집하여 학습 Dialog 개발 Intent Entity Dialog Test Iterations 문장내에서 추출해야 하는 다양한 Entity를 등록 à application에서 변수로 사용 Intent, Entity에 따라 혹은 Context, 조건에 따라 응대할 대화(응답)의 Flow를 개발 대화 테스트 38 IBM Confidential © 2019 IBM Corpora6on 22
- 23. 23 Cloud native APIs that transform voice into written text (STT) and written text into voice (TTS). Both use deep-learning AI algorithms to apply knowledge about: • grammar • language structure • voice audio signal composition ..to generate text transcriptions (STT) and voice audio with intonation (TTS). IBM Watson Speech Services Speech to Text (STT) Text to Speech (TTS)
- 24. 24 Cloud Platform Availability Watson Speech Services added ICP support to its platform in Q3 2018 and was among the first Watson service offerings* to deliver this plan option to market. ICP for Data support officially GA on January 18, 2019. *alongside the Watson Assistant and Compare & Comply teams Cloud native, behind the firewall, client managed PRIVATE Maximize on cloud agility and economics PUBLIC Value-add features plus compute and data isolation PREMIUM *STT Only. TTS in 2019. ICP for Data GA Jan 18, 2019
- 25. Discovery Service 25 Discovery Service Question : What did Donald Trump do in visit to Japan? 33 IBM Confiden+al © 2019 IBM Corpora+on 25
- 26. Watson Natural Language Understanding 다양한 글 NLU 그 글로 부터 키워드, 엔티티, 연관어, 분류체계, 감성 와 같은 정형화된 인사이트 도출 34 IBM Confiden+al © 2019 IBM Corpora+on 26
- 27. Watson Knowledge Studio for M/L annotator 실행 / 분석Rule 기반 개발 M/L 기반 개발 On-Prem Cloud Watson Knowledge Studio Content Analytics Studio Watson Explorer Watson Discovery 내부 비정형 Natural Language Understanding Annotator 개발자 Rule기반 Annotator M/L Annotator M/L Annotator M/L Annotator 검색 키워드, 엔티티, Category 실행 / 분석 데이터 분석 /마이닝 데이터 분석결과 분석사용자 어플리케이션 및 DB 외부 데이터 (뉴스, 블로그, 특허, 리포트, SNS) 데이터 연계 / API 데이터 제공 / FILE JSON JSON 35 IBM Confiden+al © 2019 IBM Corpora+on 27
- 28. Watson Personality Insight 사람이 작성한 글 Personality Insight 개인이 작성한 글을 바탕으로 5가지 성격 특성, 니즈, 가치 를 포함한 글쓴이의 성향을 분석 36 IBM Confiden+al © 2019 IBM Corpora+on 28
- 29. Visual Recognition 29 Watson Sees.. 37 IBM Confiden+al © 2019 IBM Corpora+on 29
- 30. Watson Engagement WATSON ADOPTION METHODOLOGY THINK BIG START SMALL SCALE FAST ü 현업별 현행 업무개선을 위한 Watson 활용 CASE 및 구체적인 To-be 시나리오 도출 (DESIGN THIKING 방법론 적용) ü 우선순위 검토를 통한 단기 및 중장기 JOURNEY 수립 ü 내부 READINESS 검토 및 ENABLEMENT(DATA, RESOURCE AND TECHNOLOGY) ü 연관업무를 중심으로 빠르게 확산 ü 중장기적으로 AI HUB 조직하에 (Competency Center) 확산방향에 대한 의사결정 체계화 ü 필요시 가치가 없는 프로젝트의 경우, 과감히 중단 ü 빠르게 구현à 적용à 검증할 수 있는 단기과제 선정 (DATA, 구현의 제약사항 및 Risk의 이해 필요) ü 프로젝트 추진 (MVP구현 또는 PILOT PROJECT) ü 과제는 지속적인 반복 (Iterations) 형태로 수행하며, 초기부터 사용자의 feedback을 수집/반영하여 신뢰도를 향상 IBM Confiden+al © 2019 IBM Corporation30
- 31. Q & A
- 33. Appendix
- 34. IBM Watson Assistant An enterprise artificial intelligence (AI) assistant that helps businesses enhance brand loyalty and transform their customer experiences by delivering proactive and personalized services while ensuring data privacy. An AI assistant for: Your brand: Increased engagement and loyalty Your customers: Guided learning for customized experiences Your data: Secure and personalized insights IBM Cloud / © 2019 IBM Corporation
- 35. Key Use Cases Through the Watson Assistant, IBM can decrease call center operations cost, while improving the customer experience and developing new revenue streams E.g. RBS, Orange, Autodesk, Lloyds, Sprint, Vodafone, Honda Customer Care Simplify access to common questions and tasks through enterprise channels – often HR related. E.g. Woodside, Discover, Morneau Shepell, Bradesco Employee Productivity Provide guided buying experience for prospective customers to purchase goods and services through the mobile or messaging channel of their choice E.g. GEICO, Lotte Conversational Commerce INTERNAL ONLY NOT ALL CLIENTS ARE REFERENCEABLE
- 36. Key Benefits of an Assistant Improve User Experience Respond to customers by avoiding ‘wait’ queues, make self service available 24/7 while providing a consistent brand experience to your customers. Lower Operating Costs Through Watson Assistant, IBM can decrease call center operation costs through self service and deflection of incoming queries. Generate New Revenue Streams Leverage Next Best Action to upsell and drive additional revenue. IBM Cloud / © 2019 IBM Corporation
- 37. Scalable product An enterprise can go from a single user to supporting millions of conversations across its organization with a single product. Data Control Clients maintain control and ownership of their data. IBM will not share unique insights derived from clients data without client’s agreement. Simplified tooling Allows collaboration between business users and developers to build out conversational solutions and advanced dialog flows, without needing to be an expert in Machine Learning. Deep industry expertise and partnerships e.g., health, banking, automotive, retail Premium/Dedicated Plans Provide enterprise-grade security and support, such as data isolation, end to end encryption, and support for non-regulated PII data. Key Differentiators IBM Cloud Private: On-premises plan One of the only vendors in the market that allows you to deploy and run a virtual assistant on your own infrastructure, behind your firewall. IBM Cloud / © 2019 IBM Corporation
- 38. IBM Cloud / © 2019 IBM Corporation Source: https://chatbotslife.com/know-your-intent-sota-results-in-intent-classification-8e1ca47f364c 1.AskUbuntu Corpus: 5 Intents, 162 samples 2.Web Application Corpus : 8 Intents, 100 samples 3.Chatbot Corpus: 2 Intents, 206 samples Publicly available benchmark test shows Watson Assistant on top of MSFT LUIS and Google Dialogflow (Sept 2018)