Downloaded 63 times
Uploaded bySeokJin Han
20170926 a dive into microsoft strategy on machine learning, chat bot, and artificial intelligence
I've update the original version to make it better for discussions. Today Microsoft has a very complete story around AI solutions under the name of "AI Platform" (see https://www.microsoft.com/en-us/AI/ai-platform for details). This presentation is older than that, but is talking about major area where we can benefit from AI.
20170926 a dive into microsoft strategy on machine learning, chat bot, and artificial intelligence
- 2. • What isAI and why now? • Digital Transformation and AI • Microsoft Approach to AI • One perspective on AI : Reasoning • Another perspective on AI : Understanding and Interacting • Discussions and Q&A
- 3. • What isAI and why now? • Digital Transformation and AI • Microsoft Approach to AI • One perspective on AI : Reasoning • Another perspective on AI : Understanding and Interacting • Discussions and Q&A
- 7. Businesses will requireROI from AI • Investment increased 10 times recent 5 years (2011-2016), but commercial cases are limited • Drastic changes of views last 2 years (AI: from enemies to partners) Faster development on Conversational Interface • Game-changing innovations (AI learns human languages) • Natural language search from Google and Bing, DeepText from Facebook (Personal Pattern Recognition), Changes on Chat Bots/Digital Assistants/Messenger Apps Designs evolve to increase Credibility of AI • Reflects onto AI design the knowledge on how human earns credibility between people • AI NLP integrated with Communication components such as tone, emotion, timing, visual perception, and word selection Begin discussion on how AIs will talk to each other • Protocols between AIs • How to evade collision between AI systems operating as silos • Consider collisions between AI systems of different purposes Imbedded bias will be a big blocker for AI dev • Cases from Google/Microsoft • Gender, Racial imbalance • Different sources of bias • Training data, user interactions, lack of diversity, conflicting purposes Interactions Computer – Computer Human – Computer Human – Human 5 predictions for artificial intelligence in 2017, Stuart Frankel, CEO, Narrative Science, Dec 2016
- 8. • What isAI and why now? • Digital Transformation and AI • Microsoft Approach to AI • One perspective on AI : Reasoning • Another perspective on AI : Understanding and Interacting • Discussions and Q&A
- 14. • What isAI and why now? • Digital Transformation and AI • Microsoft Approach to AI • One perspective on AI : Reasoning • Another perspective on AI : Understanding and Interacting • Discussions and Q&A
- 23. Transform data intointelligent action Intelligence Dashboard / Visualization Info Mgmt Big Data Store Machine Learning / Advanced Analytics Cortana IoT Hub Event Hub HDInsight (Hadoop and Spark) Stream Analytics Data Intelligence Action People Automated Systems Apps Web Mobile Bots Bot Framework SQL Data Warehouse Data Catalog Data Lake Analytics Data Factory Machine Learning Data Lake Store Cognitive Services Power BI Data Sources Apps Sensors and devices Data
- 24. • What isAI and why now? • Digital Transformation and AI • Microsoft Approach to AI • One perspective on AI : Reasoning • Another perspective on AI : Understanding and Interacting • Discussions and Q&A
- 25. Advanced Analytics Cycle Act:Score, Visualize Deploy Apps, Services & Visualizations Measure Preparation Modeling Feature & Algorithm Selection Model Testing & Validation Models Visualizations Ingest Profile Explore Visualize Transform Cleanse Denormalize Prepare Model OperationalizeModels Visualizations
- 29. Input1 Input2 …Actual Predicted • Classification example – Confusion Matrix
- 31. Demo Rolls-Roycecasestudy https://customers.microsoft.com/en-US/story/rollsroycestory Rolls-Roycedemo http://rolls-royce.azurewebsites.net/#/fleetlocation Solutions – PredictiveMaintenance forAerospace https://gallery.cortanaintelligence.com/Solution/Predictive-Maintenance-for-Aerospace-4 Tutorial–Simulating phenotypes fromgenomic data https://gallery.cortanaintelligence.com/Experiment/Simulating-phenotypes-from-genomic-data-2 https://github.com/Azure/Cortana-Intelligence-Gallery-Content/tree/master/Resources/Phenotype-Prediction Solutions – Vehicle Telemetry(IoT) https://gallery.cortanaintelligence.com/Solution/Vehicle-Telemetry-Analytics-9 https://docs.microsoft.com/en-us/azure/machine-learning/cortana-analytics-playbook-vehicle-telemetry
- 34. • What isAI and why now? • Digital Transformation and AI • Microsoft Approach to AI • One perspective on AI : Reasoning • Another perspective on AI : Understanding and Interacting • Discussions and Q&A
- 37. { Your Code} Direct Line Protocol REST Endpoint
- 38. • Bots areUX, Conversations as a Platform (CaaP) • Contents are important as well: From simple information delivery to actionable insights 1.Microsoft R • Statistical Analysis, Data Preparation, Predictive Modeling Big Data • Hadoop, Spark, Data Lake Analytics Machine Learning • Predictive Analysis, Deep Learning Cognitive Services • Image Recognition, Natural Language Understanding Bot Framework • Dev Framework, Different service channels Technologies around Bots
- 39. Extended Scenarios Big DataAnalytics Spark on HDInsight Data Lake Analytics Real Time Processing Stream Analytics Personalized Offer Machine Learning SQL Server R Services On-premises Integration SQL Server Data Management Gateway Visualization enabled Power BI Embedded
- 41.
- 42. Bots are communicationinterfaces with natural language processing capabilities Hi Shelly! I see you’re not satisfied. How can I help? Call Schedule Customer satisfaction call scheduledwith Shelly Smith. 08:23 AM Account Shelly Smith Primary Contact Account Filt er: All Incl ude: Related *Regarding” RecordsCase Number Title Portal timesheets Call Status Active Created on 6.14.2013 9:18 AM DFC Support CasesCase Associated View W X Y Z Contact center of the future Deepen engagement Hidden insights beyond your data Infusing business processes with intelligence Business Systems Integration Machine Learning Big Data Deepen engagement Smart recommendations, personalization, immersive experiences Infusing business processes with intelligence Dynamics CRM Hidden insights beyond your data Advanced analytics, Azure Machine Learning, data enrichment
- 44.
- 45. Democratizing AI To empowerevery person and organization to achieve more with AI
- 46. • What isAI and why now? • Digital Transformation and AI • Microsoft Approach to AI • One perspective on AI : Reasoning • Another perspective on AI : Understanding and Interacting • Discussions and Q&A Food for thoughts – How retailers can drive digital transformation with AI
- 47. Delight your customerswith personalized experiences Empower your workforce to provide differentiated customer experiences Transform your products and services Optimize your supply chain with intelligent operations Shoe style XYZ In-stock at Modern Store! BUY NOW SHOP SIMILAR CHAT Digital assistant Product expert alert It looks like Jane might need help CUSTOMER HISTORY Women’s Clothing Intelligent Customer Service Shoe style XYZ In-stock at Modern Store! BUY NOW SHOP SIMILAR CHAT Digital assistant Shoe style XYZ In-stock at Modern Store! BUY NOW SHOP SIMILAR CHAT Digital assistant Product expert alert It looks like Jane might need help CUSTOMER HISTORY Women’s Clothing Intelligent Customer Service Proposed production based on forecasted trends High-performing attributes Blue Leather Cross- body Recommendation: More blue, leather, and cross-body styles Demand forecasting 1 2 3 4 5 6 7 Green clutc Floral hand Leather cro Cross-body Demand forecasting Blue Leather Cross- body Recommendation: More blue, leather, and cross-body styles
- 48. WHERE TO BUY Howcan I help you, Jane? Can you help me find the size I’m looking for? Sure. What size are you? INTERACTIVE KIOSK MODERN STORE Shoe style XYZ In-stock at Modern Store! BUY NOW SHOP SIMILAR CHAT Shoe style XYZ In-stock at Modern Store! BUY NOW SHOP SIMILAR CHAT MODERN STORE Size 7.5 WHERE TO BUY MODERN STORE Shoe style XYZ In-stock at Modern Store! BUY NOW SHOP SIMILAR CHAT Shoe style XYZ In-stock at Modern Store! BUY NOW SHOP SIMILAR CHAT MODERN STORE Shoe style XYZ In-stock at Modern Store! BUY NOW WITH OFFER SHOP SIMILAR CHAT MODERN STORE Analyze customer behavior and sentiment and automatically deliver highly relevant product and service offers Transform the way customers begin the buying journey and use automated engines to help them take the next best step Provide customers with a digital personal assistant to guide their decision-making and shorten the conversion cycle
- 49. Product expert alert Itlooks like Jane might need help CUSTOMER HISTORY Women’s Clothing Sent to… Sales Associate Scanning for accurate pricing. . . . OPTIMAL STORE LAYOUT Children’s’Apparel Accessories Men’s Women’s Counter Men’s Women’s Counter MODERN STORE HEAT MAP OPTIMAL STORE LAYOUT Children’s’Apparel Accessories Men’s Women’s Counter Men’s Women’s Counter MODERN STORE HEAT MAP Sent to… Sales Associate Scanning for accurate pricing. . . . Optimize shelf space and ensure items are always labeled with accurate prices and promotions Develop heatmaps based on in-store customer behavior to understand which store layouts drive conversions and increase sales Alert expert staff automatically when customers have specific concerns about a product or service
- 50. Blue Leather Cross-body Recommendation: More blue, leather, andcross-body styles High-performing attributes Projected handbag demand Demand forecasting Restock now Low leather handbag stock Inventory Alert DELIVERY ROUTE OPTIMIZATION Order 10295 Optimizing local delivery route 1 2 3 4 5 6 7 Green clutc Floral hand Leather cro Cross-body Projected handbag demand 1 2 3 4 5 6 7 Green clutc Floral hand Leather cro Cross-body Blue Leather Cross-body Recommendation: More blue, leather, and cross-body styles High-performing attributes Demand forecasting DELIVERY ROUTE OPTIMIZATION Order 10295 Optimizing local delivery route Restock now Low leather handbag stock Inventory Alert Optimize order deliveries and cut costs by using customer location and predicted traffic to plan distributed order fulfillment Generate more accurate demand forecasting and pricing insights based on public and customer data automatically Automate stock replenishment processes and optimize inventory management
- 51. Based on this data,we should stock more boot colorways Based on this data, we should stock more boot colorways Spring Favorites Spring Favorites Searching Inventory Jane’s Favorites Searching Inventory Jane’s Favorites Enable customers to test, model, and customize products on the sales floor Identify customer preferences from multiple sources and match them to the most relevant piece of inventory Aggregate and analyze sentiment collected throughout the buying process to further fine-tune the customer journey
- 52.
Editor's Notes
- #5 Artificial Intelligence is term widely used nowadays but it lacks a common definition.
- #6 Wikipedia: “Machines showing capabilities that are typically associated with human intelligence”.
- #7 AI is not new. We’ve been in a journey for decades to augment human capabilities in one of those three areas. Think about the first calculators, which could do better than humans even with mechanical options like the Grant calculator in 1877. Electronics then opened a new frontier for AI. Natural experiences like Voder, a machine able to mimic human voice. Or first attempts on reasoning machines. Then computers, bringing graphical user interfaces, voice recognition or self driving cars as early as 1994. Then why are we talking about AI now? The progress all these decades has been linear, with even some setbacks. However, three things are dramatically changing the speed of innovation: Advances in AI algorithms, in particular deep learning with neural networks. Huge compute power available in the cloud. Big data, able to train these algorithms with huge sets of data.
- #36 More natural human / computer interaction Appealing. Efficient. Easy. Natural. Adapts to the user depending on circumstance Available where you are (Web, Mobile, Car, Desktop, …)
- #43 The thing is, while bots represent the future of intelligent business technologies, bots themselves are not inherently intelligent. At Microsoft, we infuse intelligence into the Bot Framework by leveraging Microsoft’s great analytics offerings, Cognitive Services, and other products that enhance the bots capabilities. Intelligent bots are critical to organizations looking to deepen engagement with users, uncover hidden insights beyond your data, and infuse processes with intelligence. <click> For example, organizations will be able to use sophisticated bots to deepen engagement with customers by meeting them where they already are across all channels to offer smart recommendations and personalized, immersive experiences. <click> The call center of the future will leverage these bots by integrating bot processes with Microsoft’s Dynamics CRM capabilities. Organizations can improve customer interactions with important contextual customer information that enables a smooth transition to a human agent when there is a need to complete more complicated transactions or inquiries. <click> Bot can also use Microsoft’s advanced analytics, Machine Learning, and data enrichment tools to draw insights from information organizational databases, as well as third-party data. What this comes down to is an improved ability to capitalize on insights about your customer from their preferences to their brand sentiment, and an enhanced ability to take informed action. <click> As you can see, intelligent bots can help your capitalize on biggest technology trends and new capabilities, including business systems integration, machine learning and big data. T: Now, let’s take a look at what is happening behind the scenes when an intelligent bot is in action. <click.
- #44 Bots can help your organization by capturing and reasoning over all data received through customer interactions. For example, here you can see that Litware Insurance has implemented a bot to assist with customer service. The bot functions externally through live chat sessions with customers to provide personalized information, track claims, and make custom recommendations – all without the help of a human agent. The bot is able to leverage all of the information gathered throughout these customer interactions internally as well. The bot is capable of identifying customer opportunities, as well as customers whose loyalty is at-risk. The bot can automatically escalate to a staff member to ensure follow-up based on unique scenarios. The bot is also able to track new information and add it to the correct customer accounts. All of these qualities enable Litware to use insight to improve the way customers are served. Transition: Let’s turn our attention to what it takes to architect an intelligent bot. <click>
- #48 In the modern retail environment, consumers are well-informed and expect intuitive, engaging, and informative experiences when they shop. <click> Retailers need solutions that can help them delight their customers with personalized experiences <click>, empower their workforce to provide differentiated customer experiences <click>, optimize their supply chain with intelligent operations <click> and transform their products and services Transition: In the following slides, we’ll take a deeper look at how retailers can make this experience a reality. <Click>
- #49 In an increasingly mobile and tech-savvy world, a “one-size-fits-all” shopping experience will no longer drive your business forward. Microsoft’s AI solutions help retailers rise to meet the challenges of a changing marketplace and deliver personalized experiences for customers. <click> Transform the way customers begin the buying journey and use automated engines to help them take the next best step Today’s digital retailer is actively using technology to transform the buying journey. In a modern retailer, solutions for real-time insights incorporate AI tools such as sentiment analysis to help them effectively recommend relevant products and services. A customer could take a picture of a product and tag it to their preferences. AI tools then generate metadata about the tagged product, such as the cut, color or style. Powered by machine learning, this data can then be matched to store inventory in order to surface targeted recommendations that delight customers. By simply taking a photo, or tagging an image on Instagram, retailers can create new access points for the customer to embark on a retail experience and meet customer where they already are—on their smart devices! Related video: https://youtu.be/hQ14_HpnBvY Customer Story: Demonstration: (link) Products: Computer Vision API, Text Analysis API, ML Server, Recommendations API <click> Analyze customer behavior and sentiment and automatically deliver highly relevant product and service offers Customers can also engage in a personalized shopping experience with tools such as interactive kiosks that recall saved products, check if a size or style is in stock, request a fitting room and more. AI capabilities enable these kiosks to help drive the customer forward in the buying journey by instantly surfacing a customer’s next best step based on behavior. For example, customer once the customer has tried on a product, the kiosk could prompt them to purchase it on their preferred device. Customer Story: Demonstration: (link) Products: Computer Vision API, Emotion API, Text Analysis API, Entity Linking API, Linguistic Analysis API, ML Server, Recommendations API <click> Provide customers with a digital personal assistant to guide their decision-making and shorten the conversion cycle Today’s on-the-go customer doesn’t have time to waste, especially when it comes to their product purchases. Personal digital assistants powered by AI are an excellent solution. Imagine if a customer could engage directly with a personal assistant chatbot on their preferred device to guide their decision-making. The bot would interact with the customer using natural question-and-answer language to determine the kind of product the customer is looking for, then provide a curated gallery of recommendations, along with cross-sell and upsell suggestions. The customer could select the products they like and ask the bot if they are in stock, and the bot would respond with inventory availability, then ask the customer if they would like to purchase the products. Not only would this approach shift strain off of live sales associates, it would shorten the conversion cycle and provide an efficient shopping experience. Customer Story: Demonstration: (link) Products: Bot Framework, Language Understanding Intelligent Service (LUIS), Computer Vision API, Text Analysis API, ML Server, Recommendations API Transition: Let’s discuss the ways that AI can empower your workforce. <click>
- #50 A majority of companies are focused on improving customer experiences. By enhancing your business’ day-to-day strategies and operations with Artificial Intelligence solutions, you can empower your workforce to deliver not only improved, but differentiated customer interactions. <click> Develop heatmaps based on in-store customer behavior to understand which store layouts drive conversions and increase sales A modern retailer utilizes digital technology to better understand customer shopping behaviors and drive conversions. Using cameras powered with AI capabilities, a retailer can track customer traffic patterns to build a heatmap of where they go and what products they’re looking at. This enables them to take the heatmap and apply advanced analytics to make store layout recommendations. By overlaying this data with POS data, retailers can further refine their models and change store layouts to increase sales Customer Story: Demonstration: (link) Products: Computer Vision API, Face API, ML Server, Recommendations API, Dynamics 365 <click> Optimize shelf space and ensure items are always labeled with accurate prices and promotions Some of biggest challenges that retailers face are missing or misplaced products on shelves and incorrect product labels. A traditional approach is to send an associate around on an hourly basis to ensure sure no one has picked up a product and put it back in wrong place. With AI tools, retailers can more effectively address these issues and free up valuable sales associate time to focus on delighting customers. Using AI, retailers can empower store cameras to scan shelves or displays and match the layout with their plan-o-grams. Missing or incorrect items would automatically be flagged, with notifications pushed to the appropriate employee for action. Customer Story: Demonstration: (link) Products: Computer Vision API, Custom Vision Service, Text Analysis API, ML Server <click> Alert expert staff automatically when customers have specific concerns about a product or service Provide excellent service when and where your customers need it by connecting customers with sales associates more efficiently using AI. AI capabilities enhance existing security cameras by taking the video feed and analyzing behavioral patterns to identify customers who may be confused and need help. Based on where they're located in the store, the tool can then send notifications to an associate with expertise in that product. For instance, if a customer is in the hardware store paint section, the tool will send a notification to the paint expert, who will approach that customer. Customer Story: Demonstration: (link) Products: Bot Framework, Computer Vision API, Face API, ML Server, Recommendations API Transition: These intelligent solutions can play a part in optimizing your supply chain. <click>
- #51 Intelligent inventory and supply chain management not only drastically reduces costs, it makes businesses more agile and efficient while improving the customer experience. <Click> Generate more accurate demand forecasting and pricing insights based on public and customer data automatically Successful retailers know that effective operations starts with the right balance between supply and demand. If too many products are ordered, they may not sell. If too few products are ordered, it impacts the customer experience. AI tools can improve a retailer’s ability to accurately forecast demand and generate pricing insights. By analyzing public data such as consumer trends and sentiment in conjunction with internal POS and engagement data, a retailer can identify and track trends in order to predict optimal inventory levels and competitive price points. Customer Story: Demonstration: (link) Products: Bing News Search, Bing Web Search, Text Analytics, ML Server, Recommendations API, Dynamics 365 <click> Optimize order deliveries and cut costs by using customer location and predicted traffic to plan distributed order fulfillment Today’s modern retailer is leveraging their customers’ location to increase profit margins and deliver excellent service. Using advanced analytics, retailers can match a customer’s location with the closest warehouse in order to cut shipping costs while maintaining optimal service levels. Machine learning allows them to aggregate this data by geography to predict future purchases and recommend how to stock each warehouse. This also enables them to make bulk shipments to locations in order to further reduce shipping costs. Customer Story: Jet.com (https://blogs.microsoft.com/transform/2015/08/03/jet-com-redefines-online-shopping-with-transparent-dynamic-pricing/#sm.00002ks3urohbdd6vfs1tjx7hl4f4) Demonstration: (link) Products: Computer Vision API, Custom Vision Service, Text Analysis API, ML Server <click> Automate stock replenishment processes and optimize inventory management Infuse intelligence in traditional inventory management and optimize processes with AI. Using cognitive tools, retailers can empower store cameras to scan shelves or displays and match the layout with their plan-o-grams. Missing items would automatically be flagged for replenishment, with notifications pushed to the appropriate employee to restock or order additional inventory. Additional automation can be employed to instantly order new inventory once stock goes pre-determined certain levels. Customer Story: Demonstration: (link) Products: Computer Vision API, Custom Vision API, ML Server, Recommendations API Transition: Now let’s talk about how AI can help retailers transform their products and services. <click>
- #52 Intelligent inventory and supply chain management not only drastically reduces costs, it makes businesses more agile and efficient while improving the customer experience. <Click> Identify customer preferences from multiple sources and match them to the most relevant piece of inventory With the myriad number of products and options available at today’s retailer, it’s easy for customers to get overwhelmed. Instead of flipping through catalogues, imagine if the customer could research their products online and pin them to a board on social media, such as Pinterest. The retailer can then pull in the customer’s research and analyze it using deep learning techniques to draw on underlying properties like what kind of materials, styles and colors the customer prefers. All these different properties can be extracted from the images pinned over time. Once they analyze it, the retailer can match it to items in their inventory and present the customer with top choices that align to their preferences. With this approach, the retailer can more effectively manage the customer journey and reduce the customer’s anxiety in the buying process. Customer Story: Demonstration: (link) Products: Bing News Search, Custom Vision Service, Computer Vision API, Bing Web Search, Text Analytics, ML Server, Recommendations API <click> Enable customers to test, model, and customize products on the sales floor Today’s customer is seeking interactivity, which means they don’t want to just see a product on a sales floor under fluorescent lights, they want to envision it in their own environment. Imagine instead if a customer could use virtual reality gear to engage with a holographic image of a product or design based on their preferences. They could freely interact with the 3-dimensional space, even changing out products for different colors or styles. At the end of their exploration, they’ve created their design exactly how they wanted it and the retailer can easily package the product selections, creating a frictionless way for the customer to say this exactly what they want. Customer Story: Demonstration: (link) Products: Computer Vision API, Custom Vision Service, Text Analysis API, ML Server, Hololens, Face API, Bing Speech API <click> Aggregate and analyze sentiment collected throughout the buying process to further fine tune the customer journey Derive insight from customer sentiment throughout the buying process by engaging AI. By analyzing public data such as product reviews and social media sentiment, in conjunction with internal POS data and store heatmaps, a retailer can better gauge product performance and further refine the customer journey. Customer Story: Demonstration: (link) Products: Bing News Search API, Bing Web Search API, ML Server, Recommendations API, Dynamics 365 Transition: Microsoft AI solutions can help retailers transform their business and enhance the customer journey. <click>