Digital Decisioning for the New Decade - 2020 and Beyond
1 like87 views
This document discusses artificial intelligence (AI) trends for 2020 and beyond in supply chain decision making. It notes that AI represents a significant business opportunity for many companies according to a recent survey. The document outlines how AI-powered decisioning can provide value in supply chains through techniques like optimization, advanced analytics, machine learning, and deep learning. It provides examples of AI solution techniques that can help with problems like demand forecasting, root cause analysis, and supplier risk segmentation. Finally, it offers tips for organizations looking to apply AI, such as identifying early adopters and building applications iteratively.
Digital Decisioning for the New Decade - 2020 and Beyond
- 1. Digital Decisioning for the New Decade: 2020 and Beyond 6th February, 2020 Andy Fox Senior Director: Opex Analytics, a LLamasoft company
- 2. Enormous Value of AI in Supply Chain Source: McKinsey Economist.com Economic-value creation from AI in the next 20 years (in $US trillion)
- 3. October 2019 Survey from MIT Sloan & BCG: 90% companies agree that AI represents a significant business opportunity for their company. 70% of companies report minimal or no impact from AI so far. Source: https://www.bcg.com/d/press/15october2019-MIT-BCG-winning-with-ai-231852
- 4. AI-powered decisioning for the supply chain
- 5. Data science & operations research experts Twenty-Three of Gartner’s Supply Chain Top 25 Make their decisions with LLamasoft technology 25% of Revenue spent annually on R&D >50% of FORTUNE 100 Rely on LLamasoft for technology enabled decisions Opex Analytics Global leader in AI-based solutions and deployment services Global team with over 750+ Employees 200+ Headquartered in Ann Arbor, Michigan, USA 20 locations around the world Supported more than 3,000 Supply Chain Analytics Initiatives 750 Global Clients for over Over $15 Billion Dollars In value identified by LLamasoft customers
- 6. AI: How We Got Here 2006/7
- 7. How Analytics Evolved Predictive Machine Learning Simulation Forecasting Alerts Descriptive Reports Drilldown Machine Learning Visualization Prescriptive Optimization Stochastic Opt Predictive Supervised learning Simulation Forecasting Alerts Descriptive Reports Drilldown Unsupervised learning Visualization Prescriptive Optimization Stochastic Optimization
- 8. AI is an Umbrella Term for a collection of algorithmic techniques OPTIMIZATION ADVANCED ANALYTICS MACHINE LEARNING DEEP LEARNING REINFORCEMENT LEARNING SIMULATION …
- 9. New Cognitive Capabilities Enable New Value TRADITIONAL IDEAS Predict Solve Puzzles SEE THE NEW IDEAS (COGNITIVE) HEAR READ ALGORITHMS DATA COMPUTING ESSENCE OF AI + + Analytics Beyond Reporting
- 10. 2020 AI TRENDS
- 11. AGI ANI Social AI Enterprise AI Flavors of AI are evolving
- 12. Organizations are understanding what AI is not AI DIGITAL AI RPA
- 13. Macro Optimization, Heuristics Micro Machine Learning Feature Driven | Very Specific | Market Driven Test Simulation, Reinforcement Learning AI Algorithms support Micro & Macro decisions Hierarchical optimization on cost vs. service Novel inventory policies Forecasting demand, delivery agents, transportation rates
- 14. Students with R or Python experience Executives realizing the need for tailored AI Platforms vs. “off the shelf” tools Algorithm Libraries More data: internal & external Rise in development of custom AI apps Supply chains are not starting from scratch with AI
- 16. 16 AI Solution Techniques for Troublesome Problems Point vs. Range Forecasting How to avoid service issues? Increasing complexity with a growing tail Unpredictable demand
- 17. 17 Outcome-based Action vs. “Looking Backwards” Point vs. Range Forecasting PREDICTION • 100s of features, mostly from Retailer data ➢ Consumption forecasts & actuals ➢ Historic orders & fulfillment ➢ Glance views & pageview out-of-stocks ➢ Customer inventory on-hand • Random Forest Quantile Regressor models potential order patterns “Feature engineering”: the art of the science
- 18. 18 Outcome-based Action vs. “Looking Backwards” Point vs. Range Forecasting OPTIMIZATION • Generates probability of not meeting demand at current inventory levels • Calculates inventory gaps • Recommends actions ➢ Stock transfers ➢ Production timelines ➢ Production quantities
- 19. ML to Derive Input Parameters for Inventory Models Point vs. Range Forecasting MACHINE LEARNING POWERED PARAMETERS OPTIMIZATION POWERED MODEL
- 20. Root Cause Detection Algorithm for Service Losses Next Generation Root Cause Analysis: Intelligent and Automated GOAL Replace manual cause coding of service losses with algorithmic approach, increasing standardization and objectivity of root cause assignment while eliminating manual planner effort APPROACH Non-linear Root Cause Analysis – Feature Creation, Scoring, & Unsupervised Learning Quality Supplier Delays National Forecasts Regional Deployment Production Attainment Production Capacity
- 21. On-Time Delivery and Chargeback Root Cause Analytics Next Generation Root Cause Analysis: Intelligent and Automated ORDER ATTRIBUTES ORDER FLOW EVENTS BUSINESS PROCEDURES CARRIER PROCEDURES TRANSIT TIMES CARRIER EVENTS FORECASTS HOURLY DAILY QUARTERLY Predicting and expediting late orders Measuring root cause of late orders Quantifying investment opportunity ANALYTICS ENVIRONMENTAL FEATURES DATA
- 22. On-Time Delivery and Chargeback Root Cause Analytics Next Generation Root Cause Analysis: Intelligent and Automated
- 23. Status quo: actions to mitigate supplier risks provide data Supplier Risk Segmentation Perform supplier audits and periodic tests Create contingency plans Collect third-party information › Supplier news stories › Geopolitical events › Natural disasters
- 24. AI Use Case: Clustering to create Supplier Segments Supplier Risk Segmentation
- 25. Applying AI to your Organization
- 26. Three use cases for data science Generate Insights One-time insight Engines Feeds enterprise systems Decision Products Business user consumer
- 27. Supply chain professionals need applications
- 28. AI departments should be “App Building Factories”
- 29. Get started with AI: our Top Five tips 1 AI Discovery and Immersion What problem areas in your organization may have new and/or different data to make better decisions? 2 Identify Early Adopters Which teams are data savvy today? Who is “off-roading” with your systems to achieve their goals? 3 Be Agile, Build Iteratively View AI initiatives as “products”, not “projects”. Start small and leverage the Early Adopters for feedback. 4 Invest in Talent and Platforms Ensure that investments in AI and data scientists are tailored to the problems you’re trying to solve 5 Keep Learning LLamasoft seminars and Analytics Academy
- 30. Continue the discussion FREE SEMINAR: Make Smarter Business Decisions with a Supply Chain Digital Twin 27th February 2020, LLamasoft Offices, Birmingham Discover how companies including AB InBev, Heinz, Tesco and Unilever are harnessing the power of a digital twin to make smarter decisions, better predict outcomes and prescribe actions, improve efficiency, performance and responsiveness, and democratise supply chain intelligence. Visit the LLamasoft booth today to collect your exclusive invitation!
- 31. Thank you