Microsoft is all about enabling partners to innovate, grow & differentiate with AI.
As Microsoft Partners we are all about enabling our clients to be the best they can be in their market segment.
Cloud computing changed the way products and services are imagined and created. It changed how software solutions were architected and enabled developers to build apps on a planetary scale. And it paved the way to democratize artificial intelligence (AI) for developers across the globe.
Microsoft Azure leads the pack in empowering developers with a spectrum of AI services. But as so often happens when we’re spoiled with choices, it becomes difficult to discern what services are the best ones to use and when. In this article, we walk through the AI services available on Microsoft Azure and provide the information and context you need to analyze your requirements and make informed decisions about which services to use based on your requirements. In some architectures, you might end up with a pipeline of multiple AI services, working in concert to achieve your goals.
Among the range of services available, Microsoft Azure provides more than 35 services targeted at AI and machine learning (ML). These services range from ready-to-use RESTful APIs to the services and SDKs that you can use to create an AI model from scratch using your own data. As we go forward, we’ll discuss these services and share code that you can try yourself. We will also explore the scenarios, requirements and target audience for these services, so you can understand when and what services make a case for themselves.
As a practicing software or ML developer/architect, you may already be aware that the terms AI and ML refer to a vast domain of software tools, frameworks, algorithms and data processing techniques. The features that are powered by AI can range from the simple—say, a personalized recommendation system—to extremely complex—such as identifying actions in videos.
The AI and ML services on Azure cater to everything under the sun and include pre-built AI services that you can integrate into your solution with a few RESTful API calls. If your specific requirement mandates a custom ML model, there are tools and services to build and use those, as well. Openness has been a core tenet of the Microsoft Azure platform, and the Microsoft AI platform allows you to use any toolset or framework that you use today, and still leverage Azure AI Services to streamline the ML workflow end-to-end.
An Overview of AI Services
For simplicity, we can arrange AI Services into three high-level categories: AI Apps and Agents, Knowledge Mining and Machine Learning. Let’s take a moment to explore these categories.
AI Apps and Agents: As a category, AI Apps and Agents are a subset of services offered by Microsoft Cognitive Services and the Azure Bot Service. Microsoft Cognitive Services are pre-built AI services that allow developers to quickly add intelligent features in their apps and services with just a few RESTful API calls. Some of these services provide customization options to suit your specification requirements, but they’re largely powered by models developed, trained and hosted by Microsoft, so you don’t really need to go through the complete process of developing a model from scratch.
Azure Bot Service is powered by the Microsoft Bot Framework platform and provides you with ways to quickly get started with functioning bots that you can extend using Microsoft Cognitive Services or other AI services from Azure. You can also integrate any AI services that you’re developing on your own. The Speech and Language Cognitive Services are also referred to as Conversational AI services in conjunction with Azure Bot Service.
Knowledge Mining: Knowledge Mining refers to a branch of Azure AI platform where intelligent information extraction is the key channel to surface insights from a corpus of structured and unstructured data. Microsoft Azure Search (with its Cognitive Skills) and Azure Form Recognizer service are two shining examples of services that are part of this category.
Machine Learning: Both the AI Apps and Agents and Knowledge Mining categories include pre-built AI services. However, Azure AI platform includes bespoke services that will make any data scientist feel at home. The Machine Learning category includes Microsoft AI Platform services to bring your ML models to life. These Azure services help you manage your ML experiments from data prep stages, to testing and managing the training, to the evaluation of runtimes. They also provide a flexible compute targets option so you can focus solely on your experiment code, rather than worrying about infrastructure, platforms and scaling. Azure Machine Learning Studio, Automated Machine Learning service (also with SDK), Azure Notebooks and Azure Machine Learning Services provide you with platform tools and services to make your experiment development more productive.
It’s one thing to review the list of available services. It’s quite another to understand how they work and interact in the context of a realistic scenario that applies to a broad range of organizations. With that in mind, we’ll describe a scenario and then use that to show how specific Azure components can address associated tasks.
For our journey today let’s start with a bit of background. Suppose you’re a lead developer, architect or AI engineer at everyone’s favorite company, Contoso. Contoso recently announced the modernization of its business processes, with one of the projects aimed to transform an expense management solution built in-house a decade ago. While the core business logic and approval workflows might not necessarily change, AI can be used to improve user interaction, data input and information processing.
To make this clearer, let’s understand the possible tasks involved in the project and then assign a specific AI service to each task for implementation.