Choose the best AI service for your needs

Choose the best AI service for your needs

 

Introduction

 

Artificial Intelligence (AI) is a category of computing that adapts and improves its decision-making ability over time based on its successes and failures. Microsoft Azure provides several AI solutions to choose from, each one depending on the problem you're trying to solve.

Tailwind Traders, a traditional brick-and-mortar retailer that has experienced explosive online sales growth, faces exciting challenges as it seeks to improve its e-commerce and service operations. Microsoft's AI services might be a good fit for one of the company's new initiatives, but Tailwind Traders needs help to better understand which product option is best for each scenario.

In this module, you'll learn about the various Microsoft AI services, and you'll analyze the decision criteria that experts use to select the right service for a specified scenario.

Learning objectives

After completing this module, you'll be able to:

• Choose the Azure AI services that best address your company's business challenges.

Prerequisites

• Familiarity with the concept of application programming interfaces, or APIs. Programmers use APIs to interact with the functionality that's contained in code libraries.
• Familiarity with the following additional concepts:
  • Web API: An API that's accessible from servers that accept requests via HTTP.
  • Web API endpoint: The location of the code library.
  • REST API: The design of the URL style that's used to expose the API's functionality.

 

Identify the product options

 

AI is a broad classification of computing that allows a software system to perceive its environment and take action that maximizes its chance of successfully achieving its goals. A goal of AI is to create a software system that's able to adapt, or learn something on its own without being explicitly programmed to do it.

There are two basic approaches to AI. The first is to employ a deep learning system that's modeled on the neural network of the human mind, enabling it to discover, learn, and grow through experience.

The second approach is machine learning, a data science technique that uses existing data to train a model, test it, and then apply the model to new data to forecast future behaviors, outcomes, and trends.

Forecasts or predictions from machine learning can make apps and devices smarter. For example, when you shop online, machine learning powers product recommendation systems that offer additional products based on what you've bought and what other shoppers have bought who have purchased similar items in the past.

Machine learning is also used to detect credit card fraud by analyzing each new transaction and using what it has learned from analyzing millions of fraudulent transactions.

Virtually every device or software system that collects textual, visual, and audio data could feed a machine learning model that makes that device or software system smarter about how it functions in the future.

Azure product options

At a high level, there are three primary product offerings from Microsoft, each of which is designed for a specific audience and use case. Each option provides a diverse set of tools, services, and programmatic APIs. In this module, we'll merely scratch the surface of the options' capabilities.

Azure Machine Learning

Azure Machine Learning is a platform for making predictions. It consists of tools and services that allow you to connect to data to train and test models to find one that will most accurately predict a future result. After you've run experiments to test the model, you can deploy and use it in real time via a web API endpoint.

With Azure Machine Learning, you can:

• Create a process that defines how to obtain data, how to handle missing or bad data, how to split the data into either a training set or test set, and deliver the data to the training process.
• Train and evaluate predictive models by using tools and programming languages familiar to data scientists.
• Create pipelines that define where and when to run the compute-intensive experiments that are required to score the algorithms based on the training and test data.
• Deploy the best-performing algorithm as an API to an endpoint so it can be consumed in real time by other applications.

Choose Azure Machine Learning when your data scientists need complete control over the design and training of an algorithm using your own data. The following video discusses the basic steps required to set up a machine learning system.

 

Azure Cognitive Services

Azure Cognitive Services provides prebuilt machine learning models that enable applications to see, hear, speak, understand, and even begin to reason. Use Azure Cognitive Services to solve general problems, such as analyzing text for emotional sentiment or analyzing images to recognize objects or faces. You don't need special machine learning or data science knowledge to use these services. Developers access Azure Cognitive Services via APIs and can easily include these features in just a few lines of code.

While Azure Machine Learning requires you to bring your own data and train models over that data, Azure Cognitive Services, for the most part, provides pretrained models so that you can bring in your live data to get predictions on.

Azure Cognitive Services can be divided into the following categories:

• Language services: Allow your apps to process natural language with prebuilt scripts, evaluate sentiment, and learn how to recognize what users want.
• Speech services: Convert speech into text and text into natural-sounding speech. Translate from one language to another and enable speaker verification and recognition.
• Vision services: Add recognition and identification capabilities when you're analyzing pictures, videos, and other visual content.
• Decision services: Add personalized recommendations for each user that automatically improve each time they're used, moderate content to monitor and remove offensive or risky content, and detect abnormalities in your time series data.

 

Azure Bot Service

Azure Bot Service and Bot Framework are platforms for creating virtual agents that understand and reply to questions just like a human. Azure Bot Service is a bit different from Azure Machine Learning and Azure Cognitive Services in that it has a specific use case. Namely, it creates a virtual agent that can intelligently communicate with humans. Behind the scenes, the bot you build uses other Azure services, such as Azure Cognitive Services, to understand what their human counterparts are asking for.

Bots can be used to shift simple, repetitive tasks, such as taking a dinner reservation or gathering profile information, on to automated systems that might no longer require direct human intervention. Users converse with a bot by using text, interactive cards, and speech. A bot interaction can be a quick question and answer, or it can be a sophisticated conversation that intelligently provides access to services.

Analyze the decision criteria

In this unit, you'll analyze the criteria that experts employ when they choose an AI service for a specific business need. Understanding the criteria can also help you better understand the nuanced differences among the products.

Are you building a virtual agent that interfaces with humans via natural language?

Use Azure Bot Service when you need to create a virtual agent to interact with humans by using natural language. Bot Service integrates knowledge sources, natural language processing, and form factors to allow interaction across different channels.

Bot Service solutions usually rely on other AI services for such things as natural language understanding or even translation for localizing replies into a customer's preferred language.

Before you jump in to build a custom chat experience by using Bot Service, it might make sense to search for prebuilt, no-code solutions that cover common scenarios. For example, you can use QnA Maker, which is available from Azure Marketplace, to build, train, and publish a sophisticated bot that uses FAQ pages, support websites, product manuals, SharePoint documents, or editorial content through an easy-to-use UI or via REST APIs.

Likewise, Power Virtual Agents integrates with Microsoft Power Platform so that you can use hundreds of prebuilt connectors for data input. You can extend Power Virtual Agents by building custom workflows with Power Automate, and if you feel that the out-of-the-box experience is too limiting, you can still build more complex interactions with Microsoft Bot Framework.

Do you need a service that can understand the content and meaning of images, video, or audio, or that can translate text into a different language?

Use Azure Cognitive Services when it comes to general purpose tasks, such as performing speech to text, integrating with search, or identifying the objects in an image. Azure Cognitive Services is general purpose, meaning that many different kinds of customers can benefit from the work that Microsoft has already done to train and test these models and offer them inexpensively at scale.

Do you need to predict user behavior or provide users with personalized recommendations in your app?

The Azure Cognitive Services Personalizer service watches your users' actions within an application. You can use Personalizer to predict their behavior and provide relevant experiences as it identifies usage patterns. Here again, you could capture and store user behavior and create your own custom Azure Machine Learning solution to do these things, but this approach would require much effort and expense.

Will your app predict future outcomes based on private historical data?

Choose Azure Machine Learning when you need to analyze data to predict future outcomes. For example, suppose you need to analyze years' worth of financial transactions to discover new patterns that could help you create new products and services for your company's clients and then offer those new services during routine customer service calls. When you're working with proprietary data, you'll likely need to build a more custom-tailored machine learning model.

Do you need to build a model by using your own data or perform a different task than those listed above?

Use Azure Machine Learning for maximum flexibility. Data scientists and AI engineers can use the tools they're familiar with and the data you provide to develop deep learning and machine learning models that are tuned for your particular requirements.

Use Machine Learning for decision support systems

The Tailwind Traders e-commerce website allows its customers to browse and purchase items that can be delivered or picked up from a retail store nearest to their location.

The Marketing team is convinced that it can increase sales dramatically by suggesting add-on products that complement the items in a shopper's cart at the point of checkout. The team could hard-code these suggestions, but it feels that a more organic approach would be to use its years' worth of sales data as well as new shopping trends to decide what products to display to the shopper. Additionally, the suggestions could be influenced by product availability, product profitability, and other factors.

The Marketing team's existing data science experts have already done some initial analysis of the problem domain, and have determined that its plan might take months to prototype, and possibly a year to roll out.

Which service should you choose?

Let's apply the decision criteria you learned about in the preceding unit to find the right option.

First, is Tailwind Traders building a virtual agent that interfaces with humans via natural language? No, it is not, so Azure Bot Service is not a good candidate for this scenario.

Second, does Tailwind Traders need a service that can understand the content and meaning of images, video, audio, or translate text into a different language? No, it doesn't, so the relevant Cognitive Services will not help the company.

Third, does Tailwind Traders need to predict user behavior or provide users with personalized recommendations? Yes, it does. However, creating recommendations based on user behavior is only part of the requirement. Tailwind Traders needs to create a complex model that incorporates historical sales data, trending sales data, inventory, and more. It's possible that the Azure Cognitive Services Personalizer service could play a role, but it couldn't handle the entire breadth of the project alone.

Fourth, will the Tailwind Traders app predict future outcomes based on private historical data? Yes, and that is why in this scenario, Azure Machine Learning is likely the best choice.

The success of this effort would depend primarily on the ability of the model to select precisely the right up-sale products to suggest to the shopper. Because the model would need to be tweaked and tuned over time, an off-the-shelf model would likely not suffice.

Finally, it sounds like the Marketing team already employs some data science experts, and the team is willing to make at least a year-long commitment to building, testing, and tweaking the models to be used.

 

Use Cognitive Services for data analysis

 

The first generation of the Tailwind Traders e-commerce website was available exclusively in English. However, when the Marketing team sponsored a demographics study for the company's brick-and-mortar locations, it found that, on average, only 80 percent of potential customers speak English. In some neighborhoods, that number falls to 50 percent. The team sees the addition of multiple languages as a wonderful opportunity to serve non-English speakers with the same online e-commerce experience as English speakers.

Which service should you choose?

As in the preceding unit, apply the decision criteria you learned about earlier to find the right option.

First, is Tailwind Traders building a virtual agent that interfaces with humans via natural language? No, it is not, so Azure Bot Service is not a good candidate for this scenario. However, should Tailwind Traders ever implement a customer service agent, it might want to consider using the Translator API to provide real-time translation to help customers who are not English speakers.

Second, does Tailwind Traders need a service that can understand the content and meaning of images, video, audio, or translate text into a different language? Yes, it does. Translating textual content from one language into another is a general purpose task that you can simplify by using the Azure Cognitive Services Translator service. The service is easy to integrate into your applications, websites, tools, and solutions. It allows you to add multilanguage user experiences in more than 60 languages, and you can use it on any hardware platform with any operating system for text-to-text language translation.

Azure Cognitive Services is likely the best option for this scenario, but let's continue applying the decision criteria to make sure.

Third, does Tailwind Traders need to predict user behavior or provide users with personalized recommendations? No, it doesn't, so the Azure Cognitive Services Personalizer is not a good candidate for this scenario.

Finally, will the Tailwind Traders app need to predict future outcomes based on private historical data? No. Although it's possible to create a Machine Learning model for multilanguage translation, it would be expensive and time consuming for Tailwind Traders to attempt to build translation models themselves. The team has neither the deep learning competency nor the linguistic data that's required to train the models.

Now that you've examined all the expert criteria, you can confidently select Cognitive Services as the best product option for this scenario.

---

Use Bot Service for interactive chat experiences

The Customer Service team has long asked for a virtual agent to handle the vast majority of questions it gets asked. No matter how prominent it makes the answers to the most frequently asked questions on the website, shoppers are impatient and perceive contact in a chat window as saving them time.

The team wants shoppers to feel as though they're interacting with a real human. When it becomes clear that the virtual agent can't provide an answer, the chat session should be transferred to a human.

Providing a virtual agent would decrease the amount of time it takes for all shoppers to receive answers. The virtual agent could answer most questions, which would free up human customer service agents to provide support for more difficult questions or thorny account-related issues.

Which service should you choose?

Once again, apply the decision criteria you're now familiar with to find the right product.

First, is Tailwind Traders building a virtual agent that interfaces with humans via natural language? Yes, it is. Azure Bot Service should be used in this scenario to implement a virtual agent chat experience. Bot Service could benefit from the information on the website's Frequently Asked Questions page, along with thousands of chat sessions that have been stored between shoppers and customer service representatives. Customer Service supervisors can test and tweak the answers to continue to refine the chat experience.

Even though you've likely found the best option for this scenario, keep applying the decision criteria to see whether any additional options might work.

Second, does Tailwind Traders need a service that can understand the content and meaning of images, video, audio, or translate text into a different language? Possibly, yes. In this scenario, Azure Cognitive Services could be used along with Bot Service to build the solution. To expedite implementation, the developers could explore using prebuilt solutions, such as QnA Maker (part of Cognitive Services) or Power Virtual Agents. Also, any Azure Bot solution would likely implement several Azure Cognitive Services, such as Language Understanding (LUIS) and possibly Translator, to translate from the shopper's language to English and back again.

Third, does Tailwind Traders need to predict user behavior or provide users with personalized recommendations? No, it doesn't. Azure Cognitive Services Personalizer is not a good candidate for this scenario.

Finally, will the Tailwind Traders app need to predict future outcomes based on private historical data? No. Although Tailwind Traders does have historical data to feed into a model, which would make it possible to use Azure Machine Learning to create a chat solution, another option is already tailored for the chat bot experience.

---

 

Summary

Our goal in this module was to help Tailwind Traders explore several AI service offerings from Azure that it can apply to various business opportunities.

You identified a few product options and their capabilities, including Azure Bot Service, Azure Cognitive Services, and Azure Machine Learning. You analyzed certain decision criteria to help yourself choose one option over another depending on the scenario. Then you applied those decision criteria to three Tailwind Traders initiatives, helping the company find the best service option for each scenario.

Without AI services, Tailwind Traders would spend more time and effort on manual tasks, respond to customers less quickly, offer weak product recommendations, and be unable to fully support customers who speak languages other than English.

AI is one focus that could transform every area of a business. Such transformation is limited only by the creativity and imagination of the organization.