Enabling AI governance through AI business Description Standard

E7d6e390a90513756419be75a43609ca?s=47 finid
June 29, 2019

Enabling AI governance through AI business Description Standard

What questions should business people ask before investing in AI? Given the inherent complexity and continuous evolution of AI technologies, descriptions of AI-enabled products either avoid discussing the specific nature of AI capabilities, or overwhelm the reader with a barrage of technical terms. In such situation, there is a need for an open standard for describing AI capabilities in terms that would enable an effective evaluation of such capabilities from the business perspective.

The purpose of this presentation is to propose an AI business description framework, a set of principles for developing an open AI business description standard. The standard is proposed to include a set of guiding questions that business decision makers should ask when evaluating AI-enabled products, as well as a set of standards for evaluating answers to such questions. The presentation will also propose a set of processes for engaging AI and business communities in the development and maintenance of the standard.

E7d6e390a90513756419be75a43609ca?s=128

finid

June 29, 2019
Tweet

Transcript

  1. None
  2. Big Data & AI Conference Dallas, Texas June 27 –

    29, 2019 www.BigDataAIconference.com
  3. WHY WE NEED AN AI FOR BUSINESS FRAMEWORK By Anna

    Sidorova, Ph.D. © 2019 Anna Sidorova. All rights
  4. ABOUT ME • Anna Sidorova, Associate Professor, ITDS, UNT •

    Research in AI strategy and governance • Teach graduate course on AI in Business • Founder of AI4B Open Project • Contact me • Anna.Sidorova@unt.edu • https://www.linkedin.com/in/annasidorov a42/ • Anna.sidorova100@gmail.com © 2019 Anna Sidorova. All rights reserved.
  5. AGENDA • Motivation • About the AI4B Framework and Standard

    • AI4B Guiding Principles • AI4B Framework Architecture and the restaurant metaphor • About the AI4B Open Project • Closing remarks © 2019 Anna Sidorova. All rights reserved.
  6. Industry reports AI, ML and analytics platforms AI-enabled IT solutions

    Employees © 2019 Anna Sidorova. All rights reserved. https://www.bcg.com/en-us/publications/201 9/most-innovative-companies-artifcial-int elligence-innovation-machine.aspx https://www.ibm.com/watson/ai-stories/ www.h2o.ai Big data, machine learning, Scikit Learn, R, AWS, Spark, Hadoop, Splunk, Pytorch, Tensorfow Big data, machine learning, Scikit Learn, R, AWS, Spark, Hadoop, Splunk, Pytorch, Tensorfow
  7. © 2019 Anna Sidorova. All rights reserved.

  8. MOTIVATION FOR THE PROJECT AI technologists and business executives do

    not speak the same language Most organizations lack adequate processes for AI evaluation and governance Proliferation of AI solutions, vendors and projects Need for a common reference framework for describing AI in business terms Organizations can only derive value from AI if AI embedded into business processes © 2019 Anna Sidorova. All rights reserved.
  9. AI4B PROJECT INSPIRATION • CRISP-DM • Cross-industry standard practices for

    data mining • COBIT 5 • Control objectives for Information Technology • The Open Group Architecture Framework • A set detailed guidelines for enterprise architecture development and management • The Agile manifesto • A set of principles for agile software development • AI ethics frameworks and guides • World Economic Forum • IDC • AI-based automation framework • ( https://blogs.idc.com/2019/01/09/idcs-ai-based-automation-evolut ion-framework-a-new-way-to-think-about-ai-automation/ ) • Google • Responsible AI practices ( https://ai.google/responsibilities/responsible-ai-practices/) • Amazon • AWS Well-architected framework ( https://aws.amazon.com/architecture/well-architected/) • Typical ML workfow • IBM • The cognitive enterprise framework ( https://www.ibm.com/downloads/cas/GVENYVP5) © 2019 Anna Sidorova. All rights reserved.
  10. AI FOR BUSINESS FRAMEWORK AND STANDARD What? A tool for

    translating AI capabilities into a value proposition Why? To enable the business community to ask the right questions about AI products, and to enable the AI community to ofer answers needed by business decision makers How? The AI for Business framework will be developed through an on-line open collaboration project. Who? Business and AI professionals who are passionate about enabling efective communication between AI and business communities. Contact us at anna.sidorova100@gmail.com AI for Business Framework and Standard Version 0.1 AI for Business Open Project Why now? As the number of products marketed as AI proliferates, business decision makers need a framework with which to evaluate them and a standard by which to compare them. Without such standard, failures of individual AI products will negatively impact trust in all AI. © 2019 Anna Sidorova. All rights reserved.
  11. AI4B KEY USE CASES EVALUATION OF AI PRODUCTS AND SOLUTIONS

    EVALUATION OF AI VENTURES AND STARTUPS AI GOVERNANCE FOUNDATIONS • Does this AI product/solution meet my business needs? • Is the product/solution technologically sound? • How does the product/solution compare to competitors? • Does the venture address an existing business need? • Is the venture leveraging a promising AI technology • Is the technical architecture adequate? • How do we ensure that organizational AI initiatives promote stakeholder objectives? • How do we ensure decision transparency in the age of AI? © 2019 Anna Sidorova. All rights reserved.
  12. AI4B FRAMEWORK BUILDING BLOCKS • Guiding principles • Example: Business

    Alignment • Inspiration: Agile Manifesto, COBIT, TOGAF • Reference architectures • Example: AI Components Layers • Inspiration: Zachman Architecture, TOGAF • Reference processes • Example: CRISP-DM • Inspiration: ITIL, APQC • Best practices • Example: TBD • Inspiration: ITIL, DataOps • Tools and templates • Example: TBD • Inspiration: strategy maps, SWOT © 2019 Anna Sidorova. All rights reserved.
  13. AI4B GUIDING PRINCIPLES

  14. ON THE IMPORTANCE OF PRINCIPLES “Every day each of us

    is faced with a blizzard of situations we must respond to. Without principles we would be forced to react to all the things life throws at us individually, as if we are experiencing each of them for the frst time. If instead we classify these situations into types and have good principles for dealing with them, we will make better decisions more quickly and have better lives as a result” Ray Dalio, Principles, 2017 © 2019 Anna Sidorova. All rights reserved.
  15. GUIDING PRINCIPLES 7. Knowledge management 8. Life-cycle management 9. Complexity

    management 4. Beneft transparency 5. Cost transparency 6. Operational transparency 1. Business alignment 2. Internal alignment 3. Technology alignment © 2019 Anna Sidorova. All rights reserved.
  16. PRINCIPLE 1: BUSINESS ALIGNMENT – AI INITIATIVES WILL LEAD TO

    VALUE CREATION IF THEY ARE DRIVEN BY STRATEGIC BUSINESS OBJECTIVES, INTEGRATED WITH BUSINESS PROCESSES AND EVALUATED IN RELATION TO ORGANIZATIONAL PERFORMANCE METRICS. PRINCIPLE 2: INTERNAL ALIGNMENT – SUCCESSFUL DEVELOPMENT AND DEPLOYMENT OF AI CAPABILITIES REQUIRES AN INTERNALLY CONSISTENT AI ECOSYSTEM, INCLUDING AI INFRASTRUCTURE, AI FRAMEWORKS AND PLATFORMS, AI MODELS AND ALGORITHMS, DATA SOURCES AND AI TALENT POOL. PRINCIPLE 3: TECHNOLOGY ALIGNMENT – SUCCESSFUL DEPLOYMENT OF AI CAPABILITIES REQUIRES HIGH LEVEL OF DIGITIZATION OF ASSOCIATED BUSINESS PROCESSES. © 2019 Anna Sidorova. All rights reserved.
  17. PRINCIPLE 4: BENEFIT TRANSPARENCY – CLEAR ARTICULATION AND ACCURATE MEASUREMENT

    OF BUSINESS BENEFITS IS CRITICAL TO THE SUCCESS OF AI INITIATIVES. PRINCIPLE 5: COST TRANSPARENCY – CLEAR ARTICULATION AND QUANTIFICATION OF ALL COSTS ASSOCIATED WITH THE DEVELOPMENT AND DEPLOYMENT OF AI CAPABILITIES IS CRITICAL TO DERIVING BUSINESS VALUE FROM AI. PRINCIPLE 6: OPERATIONAL TRANSPARENCY – ABILITY OF BUSINESS DECISION MAKERS TO UNDERSTAND AND MONITOR THE LOGIC AI CAPABILITIES IS CRITICAL TO THE SUCCESS OF BUSINESS INITIATIVES. © 2019 Anna Sidorova. All rights reserved.
  18. PRINCIPLE 7: KNOWLEDGE MANAGEMENT – DEVELOPMENT AND DEPLOYMENT OF AI

    CAPABILITIES IS ASSOCIATED WITH CREATION OF NEW TYPE OF ORGANIZATIONAL KNOWLEDGE. THIS KNOWLEDGE NEEDS TO BE CLEARLY IDENTIFIED, AND PROCESSES FOR MANAGING SUCH KNOWLEDGE NEED TO BE ESTABLISHED. PRINCIPLE 8: LIFECYCLE MANAGEMENT – ORGANIZATIONS SHOULD HAVE PROCESSES IN PLACE FOR MANAGING AI CAPABILITIES THROUGH ALL STAGES OF LIFECYCLE, FROM DEVELOPMENT, THROUGH DEPLOYMENT, OPERATIONS AND DECOMMISSIONING. PRINCIPLE 9: COMPLEXITY MANAGEMENT – DEVELOPMENT AND DEPLOYMENT OF MULTIPLE AI CAPABILITIES MAY LEAD TO INTERACTION EFFECTS. ORGANIZATIONAL SHOULD ATTEMPT TO ANTICIPATE AND MITIGATE SUCH INTERACTION EFFECTS BY MONITORING AND MANAGING THE COMPLETE AI CAPABILITY ECOSYSTEM. © 2019 Anna Sidorova. All rights reserved.
  19. THE AI4B REFERENCE ARCHITECTURE: THE RESTAURANT METAPHOR

  20. Business Capabiliti es Platforms, Environme nts Models and algorithm s

    AI Capabiliti es Technical Infrastruct ure Loan risk assessment Credit card fraud detection Banking customer service Contract governance AI capabilities integrated within a business process: One or more AI capabilities + non-AI IT + business process NLP: e.g. speech-to-text, entity detection, translation, topic identifcation Image processing: Image classifcation, object detection, face matching Forecasting, anomaly detection, recommender systems Search, optimization, generation (images, text) Models trained to address a well-defned business problem: Model type + data + training process Traditional ML: Linear regression, logistic regression, decision trees, SVM Ensemble learning: Random forests, Gradient boosting, XGBoost Deep learning (DNN): Fully connected, CNN, RNN (LSTM, GRU) Other: evolutionary (genetic) algorithms, GANs, unsupervised learning Model forms and associated ML algorithms: Model structure + settings + training algorithm + optimizer OSS Libraries and APIs: Python Numpy, Scipy, Scikit Learn, R, Keras Proprietary platforms: SAS Vyia, H2O.ai, AWS SageMaker Deep learning frameworks: Tensorfow, Pytorch, Theano, CAFFE, MXNet Data processing tools: Spark, Hadoop, AWS Kinesis SW used to develop, train and deploy the models: Languages + libraries + platforms + frameworks Development, Training, Inference Cloud: AWS, IBM, Google, MS Azure Hardware: High Performance Clusters (HPC), IBM Power AI Components: CPU, GPU, TPU, FPGA Underlying storage and compute infrastructure: Compute + memory + storage + parallelization © 2019 Anna Sidorova. All rights reserved.
  21. THE RESTAURANT METAPHOR © 2019 Anna Sidorova. All rights reserved.

  22. INFRASTRUCTURE (GPU, TPU, Cloud, HPC) FRAMEWORKS, PLATFORMS, TOOLS (Python, Scikit

    Learn, R, Tensorfow, Pytorch, Keras, SAS Viya) MODELS AND ALGORITHMS (Regression, SVM, Decision trees, XGBoost, DNN, CNN, LSTM, NAS) AI CAPABILITIES (Chatbot, sales forecast, risk analysis, emotion identifcation, machine translation) BUSINESS CAPABILITIES (Customer service, supply chain management, loan approval process) DATA (OLTP, streaming, social media, IoT, external) TALENT (Data scientists, data engineers, ML engineers, cloud architects, developers) © 2019 Anna Sidorova. All rights reserved.
  23. THE AI4B PROJECT: NEXT STEPS

  24. NEXT STEPS • Iteratively refne AI4B guiding principles • Iteratively

    refne AI4B reference architecture • Identify and defne reference AI4B processes • E.g. develop AI strategy, plan AI capabilities, develop AI models, etc. • Inspiration: ITIL 4. APQC Process Classifcation Framework, CRISP-DM Process • In relation to AI4B processes, identify and document best practices and their sources • In relation to AI4B best practices, identify and document available templates and tools © 2019 Anna Sidorova. All rights reserved.
  25. LEARN Lessons learned from the applications are translated into wishlist

    for future releases of the AI4B Framework and Standard APPLY Elements of the AI for Business Framework and Standard are applied in practice by the members of the community INTEGRATE Contributions are reviewed and consolidated into and integrated AI4B Framework and Standard, and shared with the community CONTRIBUTE Individual contributions from business and AI technical community are received and shared AI for Business (AI4B) Open Project AI FOR BUSINESS (AI4B) OPEN PROJECT What is the goal of the project? AI for Business Open Project aims to develop the AI for Business Framework and Standard through open collaboration among business and AI professionals. Why open collaboration? An open collaboration format optimal for ensuring:  Open access to AI for Business Framework and Standard  Broad engagement from business and AI communities  Continuous updates and improvements of the AI for Business Framework and Standard through an iterative Contribute, Integrate, Apply, Learn © 2019 Anna Sidorova. All rights reserved.
  26. WWW.AI4B.NET © 2019 Anna Sidorova. All rights reserved.

  27. WITH CONTRIBUTI ONS AND INSPIRATION FROM Alex Pettit, Ph.D. (Oregon,

    USA) Raj Muppala (London, UK) Steve Ardire (Washington, USA) Chris Tyler (Texas, USA) Emrah Gozcu (London, UK) David Cordeiro (Texas, USA) Andrew Borges (London, UK) Yasemin Tarakci (Texas, USA) © 2019 Anna Sidorova. All rights reserved.
  28. THANK YOU.

  29. RESOURCES • TOGAF Principles: http://pubs.opengroup.org/architecture/togaf8-doc/arc h/chap29.html#tag_30_06 • COBIT 5: https://www.isaca.org/ecommerce/Pages/Cobit5-Downloa

    d-Registration.aspx?utm_referrer= • Agile Manifesto: https://agilemanifesto.org/ © 2019 Anna Sidorova. All rights reserved.