Azure AI is driving innovation and improving experiences for employees, users, and customers in a variety of ways, from increasing workday productivity to promoting inclusion and accessibility. The success of Azure AI—featuring Azure Cognitive Services, Azure Machine Learning, and Azure OpenAI Service—is built on a foundation of Microsoft Research, a wide range of Azure products that have been tested at scale within Microsoft apps, and Azure customers who use these services for the benefit of their end users. As 2023 begins, we are excited to highlight 10 use cases where Azure AI is utilized within Microsoft and beyond.
#1: Speech transcription and captioning in Teams
Speech transcription and captioning in Microsoft Teams is powered by Azure Cognitive Services for Speech. The Speech-to-Text service transcribes over 54 million hours of meetings in real time and captions 6.8 million characters each month. Microsoft achieved human parity in conversational speech recognition when it reached an error rate of 5.9 percent. The word error rate of professional speech transcribers is 5.9 percent. Twitter and Swedish TV are two customers using Azure AI to caption speech for accessibility.
#2: Content and design production in PowerPoint
Content and design production in PowerPoint, which has been used by nearly two billion Designer slides since its launch in 2015, is powered by Azure Cognitive Services and Azure Machine Learning, and their MLOps capabilities. These tools build models faster and at scale, replacing local development, and recommend background images and videos. The 2022 launch of DALL-E 2 in the Azure OpenAI Service improves the capabilities of this feature. Up to 4.1 million slides are created with Designer each day, and the Designer team regularly retrains models and experiments with model optimization to provide optimized content recommendations. Companies like Polaris encouraged employees to use Designer to share highlights from the year during the pandemic, when they shifted to virtual company events.
#3: Biometric detection and identity verification in Windows Hello
Biometric detection and identity verification in Windows Hello, which has been deployed to over 100,000 Microsoft employees, is powered by Azure Cognitive Services for Vision. Outside of Microsoft, Uber uses the Cognitive Services Face API, which powers Windows Hello, to ensure that drivers using the app match the account on file for fraud prevention and to increase peace of mind for both drivers and riders. This extra verification step is fast, works on all smartphones, and can scale to over one million driver-partners. Home healthcare and hospice care provider Amedisys also implemented Windows Hello for Business to improve its security strategy and patient care support.
#4: Personalized recommendations in Xbox
Personalized recommendations in Xbox powered by Azure Machine Learning and Personalizer have been rolled out to global audiences. In Microsoft Docs, referral traffic from content recommendations generate twice as many page views compared to other sources of traffic. The NBA uses Personalizer to deliver personalized content to fans at an individual level, with nearly 10 billion interactions across various delivery channels each day.
#5: Content reading and writing experiences in Edge and Microsoft365
Content reading and writing experiences in Edge and Microsoft 365 are powered by Azure Cognitive Services for Speech and Applied AI Services. Immersive Reader, which is used by over 15 million users on applications such as Teams, Word, OneNote, and Outlook, allows users to read text with an optimized layout and helps them to read 10 percent faster with 50 percent fewer reading errors. USA Today offers premium subscribers the option to “Hear This Article” using the Cognitive Services Speech-to-Text, which powers the Read Aloud feature in Immersive Reader.
#6: Image captioning and alt text in LinkedIn
Image captioning and content accessibility on LinkedIn are powered by Azure Machine Learning, Cognitive Services for Vision, and Azure OpenAI Service. Image classification and tagging, including the automatic generation of captions and alt text for images on LinkedIn, saves users time and improves accessibility for the community. In the manufacturing industry, AI-driven image classification can also be used to detect quality issues on assembly lines, reducing manual effort and improving the efficiency and quality of the manufacturing process. Microsoft achieved human parity in image captioning in 2020.
#7: Contextual Assistance in Outlook
Contextual assistance in Outlook, powered by Azure Machine Learning, has increased customer engagement. Meeting Insights, which is shown in over 40 percent of opened meetings on supported Outlook clients, has customers reporting that two out of three suggestions are useful. The learnings from these Meeting Insights models are being used to integrate other intelligent content recommendation features, such as “Suggested Attachments” and “Suggested Reply with File,” into Outlook.
#8: Conversational Intelligence in Viva Sales and Dynamics
Viva Sales and Dynamics use conversational intelligence powered by Azure Machine Learning and Cognitive Services for Speech & Language. Viva Sales helps sellers by providing real-time business context from their CRM system, facilitating knowledge sharing and collaboration, and offering AI-powered insights to improve customer engagements. Lufthansa Cargo uses Dynamics 365 Sales, Customer Service, and Customer Voice to centralize customer information and related shipments in one location.
#9: Assisted Programming and Content Generation in GitHub and Power Platform
Assisted programming in GitHub and Power Platform is powered by Azure Machine Learning and Azure OpenAI Service. GitHub Copilot has developers accepting almost 25 percent of all completions shown by the tool. Azure Speech also powers the Voice to Code experience for GitHub Copilot. The Power Platform team also realized that the ability to translate conversational language into code could make the code more accessible. CarMax used Azure OpenAI Service to help summarize 100,000 customer reviews into short descriptions, saving the company over two years of manual work.
#10: Editorial Assistance in Microsoft 365
Editorial assistance in Microsoft 365 powered by Azure Machine Learning, is available through Microsoft Editor, which gives writers intelligent tools to improve their writing in documents, emails, and posts across the web. The tools are available in Word, Outlook, and the Edge Browser, and are expanded to Outlook.com, Outlook for the web, and browser extensions for Microsoft Edge and Google Chrome in over 20 languages.
Looking ahead into 2023 with Azure AI
The ten scenarios above represent just a slice of how Azure AI helped Microsoft users and customers do more with less in 2022. We look forward to seeing how organizations utilize AI to transform business, employee, customer, and end-user experiences around the world in 2023.
Learn more
Discover more use cases, solutions, research, and customer inspiration with Azure AI.
Originally posted on January 23, 2023 @ 4:57 pm