At Build 2021, Microsoft’s developer conference, the company announced additions and improvements to its family of AI services available in Microsoft Azure. The company says they’re aimed at “significantly” reducing time to value in modernizing common business processes.

Enterprise use of AI grew a whopping 270% over the past several years, Gartner recently reported, while Deloitte says 62% of respondents to its corporate October 2018 study adopted some form of AI, up from 53% in 2019. A more recent survey published by KPMG suggests a large number of organizations have increased their investments in AI during the pandemic, to the point that executives are now concerned about moving too fast.

“Over 80 of the Fortune 100 are using Azure AI today, and of course 95% are using Azure across the landscape,” Eric Boyd, CVP of AI Platform at Microsoft, told VentureBeat in an interview. “We’ve seen how they’ve been able to really leverage AI and machine learning to help adapt over the past year. The pandemic has accelerated the digital transformation for so many companies — things that they’d planned to get done over five years got done in months.”

Azure Applied AI Services

Microsoft is folding a segment of its Azure AI services under a new brand, Azure Applied AI Services, which encompasses Azure Cognitive Search, Azure Form Recognizer, and Azure Immersive Reader, in addition to newer offerings like Azure Bot Service, Azure Metrics Advisor, and Azure Video Analyzer. These services build on the cognitive APIs from Azure Cognitive Services and Azure Machine Learning, Boyd says, supporting the development of AI technologies by providing task-specific AI and business logic.

The reorganization is intended to make it easier for customers to find solutions to common processes. Customers can customize services and extend them with their own models from Azure Machine Learning, but Microsoft created Azure Applied AI Services to target the most popular scenarios. For example, Azure Form Recognizer builds on optical character recognition, a computer vision technology that recognizes text.

“We see AI continuing to move into the mainstream, and as a result of that, we’re responding by listening to our customers and working with them to build naturally applied AI services to simplify the way they think,” Boyd said.

Starting this week, Azure Bot Service will offer a visual authoring canvas featuring extensible, open source tools. Building on the core speech and language technologies that power Azure Cognitive Services, such as Language Understanding, QnA Maker, Speech to Text, and Text to Speech, the canvas will let developers add speech and telephony capabilities and test, debug, and publish bots across multiple channels with minimal code changes.

Landing with the Azure Bot Service upgrade is Azure Metrics Advisor, which ingests time series data, using machine learning to automatically identify anomalies from sensors, products, and business metrics. Using Metrics Advisor, customers can monitor the performance of sales to manufacturing operations, benefiting from diagnostic insights.

According to Boyd, Azure Metrics Advisor grew out of work done for Microsoft search engine Bing to detect deviations from normal operations, like a sudden drop in ad revenue. Microsoft made the technology public through Anomaly Detector, a service in Azure Cognitive Services, but Azure Metrics Advisor is tailored to business use cases.

“Metrics Advisors is a solution that helps customers automatically find anomalies in products and business metrics and then sends alerts, providing root cause analysis,” Boyd said. “Wrapping all of those things together makes it much more simple for customers to deploy.”

Azure Video Analyzer, another new service in preview, similarly looks for anomalies using AI. It brings together Computer Vision from Azure Cognitive Services, an automatic captioning model, and capabilities for integrating closed-circuit video feeds and video management systems, helping developers build AI-powered video analytics from stored and streaming footage.

Boyd says Video Analyzer can be used for workplace safety, in-store experiences, digital asset management, content monetization, and more. For example, Lufthansa Group consulting company ZeroG tapped the service to develop Deep Turnaround, which generates automatic timestamps and issues alerts when planes aren’t unloaded, refueled, cleaned, restocked, and reloaded on time. When a fuel truck arrives later than predicted, Deep Turnaround alerts turnaround coordinators and other ground operations personnel, kicking off a hunt for a solution that prevents a delay like dispatching a second fuel truck to the plane.

Azure Cognitive Services

Beyond Azure Applied AI Services, Microsoft announced enhancements for products available through Azure Cognitive Services, including Document Translation and Text Analytics for Health.

Document Translation, a feature of Translator in Azure Cognitive Services announced in preview in February, is now generally available, enabling developers to translate documents while preserving the structure and format of the original document. Document Translation is designed to help enterprises and translation agencies that require the translation of complex documents into one or more languages.

In a related development, Text Analytics for health is now generally available with Text Analytics in Azure Cognitive Services. It allows developers to process and extract insights from unstructured medical data, including doctors’ notes, medical journals, electronic health records, clinical trial protocols, and more.

Another new feature of Text Analytics is Question Answering. Now in preview, Question Answering helps users find answers from a passage of text without saving or managing any data in Azure.

AI model deployment

Microsoft also announced two new machine learning capabilities to help customers accelerate the deployment of AI models. The first is Azure Machine Learning managed endpoints, an in-preview feature in Azure Machine Learning that supports the building and productization of machine learning models. Azure Machine Learning managed endpoints automate the creation and management of compute infrastructure, including updates and security, with out-of-the-box infrastructure monitoring and log analytics tools.

Azure Machine Learning managed endpoints are already being used to serve the OpenAI GPT-3 model, one of the world’s largest natural language models, in Power Apps, Microsoft says.

Arriving alongside Azure Machine Learning managed endpoints is PyTorch Enterprise on Azure, which gives Microsoft Premier and Unified Support for Enterprise customers additional benefits, including prioritized requests, hands-on support, and solutions for hotfixes, bugs, and security patches. Microsoft says it’s collaborating with PyTorch, Facebook’s machine learning framework project, to launch a new initiative, PyTorch Enterprise Support Program, that will provide PyTorch users with a more reliable production experience. Microsoft customers with Premier and Unified support using PyTorch are automatically eligible for PyTorch Enterprise and can request hotfixes in the long-term support version of PyTorch.

“We continue to focus on everywhere from the high-end developer to people who are trying to snag particular spaces,” Boyd said. “We really want to work across the spectrum and put a lot of effort into making the Azure AI platform extensible.”