AI in action: 5 real-world intelligent apps you can build on Azure

Artificial intelligence () is transforming the way we interact with technology, data, and each other. AI can help us create more engaging, personalized, and efficient experiences for customers and employees. To stay ahead in the era of AI, businesses are embracing intelligent applications across industries to take advantage of new opportunities. How can you use AI to modernize and build new digital products in your field? In this blog post, we dive into 5 types of AI-powered apps you can create with Azure's generative AI technology, cloud-scale data, and modern application platforms. We look at how industry leaders are modernizing and innovating with AI, present sample architectures, and offer solution accelerators to help you kickstart your own app development.

What are intelligent apps?

Intelligent, or AI-powered, apps leverage AI models to enhance or automate some aspects of their functionality. AI-powered apps can leverage various capabilities, such as natural language processing, computer vision, speech recognition, machine learning, and generative AI, to provide more intelligent and responsive experiences.

You can achieve tangible and meaningful business outcomes by infusing AI into your apps. A Forrester study found that organizations using Azure managed app, data, and AI services can ship applications up to 1.5 months faster. Organizations also see an increase of up to 25% in developer productivity, allowing development teams to focus less on mundane work, and more on innovation. Azure technologies are also helping organizations reduce app downtime by up to 25%, enabling better performance and reduced risk.

Here are 5 types of apps customers are building on Azure to take advantage of these and other business benefits.

Connected products

Connected smart products interact with other devices and systems through the internet. They utilize sensors, processors, services, and communication hardware which collect data, analyze it, and take actions autonomously. AI-enhanced connected products use data gathered from devices to inform intelligent models which can adapt to user preferences and provide personalized experiences. Examples include:

  • Smart home devices: thermostats, lighting systems, and security cameras.
  • Wearable health trackers: Smartwatches, fitness trackers, and medical devices.
  • Connected cars: GPS systems, real-time traffic updates, and voice control.
  • Industrial IoT: Remote monitoring and asset tracking.
  • Intelligent retail: Smart shelves that track inventory and self-checkout systems.
  • Supply chains: Using IoT data to detect inefficiencies and forecast demand.

Let's look at a real-world example. Mercedes-Benz built their connected car platform on Azure using cloud-native services and generative AI. They first modernized their platform using Azure Cosmos DB and Azure Service (AKS), allowing them to quickly update and release new features, while preserving the quality and security of vehicle data. Then they brought in generative AI capabilities to transform their driver experience, creating a voice assistant that understands voice commands and engages in interactive conversations.

Build your own solution

Azure's modern application platform of AI, apps, and database services leverages cloud-scale data, agile development methods with , and pretrained and responsible AI needed for today's connected smart products. Here's a sample architecture for an AI-powered industrial IoT solution for predictive maintenance, that uses Azure AI Services, Azure Cosmos DB, and AKS.

Solution architecture for AI-powered IoT maintenance application.Solution architecture for AI-powered IoT maintenance application.

This example shows an IoT platform in Azure that helps with manufacturing data analysis, and service and maintenance issues. The Intelligent Maintenance Guide uses generative AI and chat features for real-time alerts and guidance on resolve maintenance issues. The application runs on AKS and uses Azure Cosmos DB for machine-specific status and alert data, as well as chat history. Azure AI Search acts as the search and retrieval system, surfacing relevant data to Azure OpenAI Service.

Additional resources:

Transaction processing at scale

Processing billions of transactions daily is the new normal across industries, from e-commerce to healthcare. Transaction processing at scale (TPaS) refers to handling a high volume of transactions swiftly, accurately, and reliably. AI-based TPaS can analyze transactions against business rules, detect anomalies, and identify potential fraud. GenAI can be used to add natural language UX interactions and data summarization for users. Scenarios where AI-powered TPaS is used include:

  • E-commerce/retail: Customer orders, payments, and buy online, pick up in store (BOPIS).
  • Banking and finance: Deposits, withdrawals, transfers, and credit card transactions.
  • Travel and hospitality: Reservations, ticketing, and payment processing.
  • Healthcare: Appointment scheduling, billing, and insurance claims.
  • Supply chain and logistics: Inventory management, shipping transactions, and tracking of goods.
  • Digital advertising: Ad placements, impressions, clicks, and conversions.
  • Fraud/anomaly detection: TPaS with AI can also be used to identify patterns, anomalies, or suspicious activities in real-time across vast amounts of transactions.

In the travel industry, American Airlines modernized its Customer Hub app on Azure, allowing it to handle massive volumes of transactions including more than 16 million real-time messages and 17.4 million service calls per day. On top of this, American Airlines is applying AI and machine learning to company operations, helping reduce taxi time and provide real-time information to maintenance personnel and other airline staff.

An example of fraud detection, Manulife migrated to the cloud and took advantage of several Azure AI services, including Azure Document Intelligence, and Azure Machine Learning to help with data correlation and fraud detection. As part of their migration, Manulife used services such as AKS and Azure SQL Managed Instance to save on costs, get to market faster, and respond to customer demands more readily.

Build your own solution

Here is an architecture example for a real-time payments and transactions application leveraging Azure OpenAI Service, Azure Cosmos DB, AKS, Azure Functions, and Azure Event Hubs:

Azure architecture of payment and transaction processing at scale.Azure architecture of payment and transaction processing at scale.

The scenario involves members who have accounts, each account with corresponding balances, overdraft limits, and credit/debit transactions. Transaction data is across multiple geographic regions for both reads and writes, while maintaining consistency. Updates are made efficiently with the patch operation. Business rules govern if a transaction is allowed. An AI powered co-pilot enables agents to analyze transactions using natural language.

Here's another reference architecture for real-time fraud detection:

Detect fraud in real-time with Azure Cosmos DB, AKS, and Azure AI services.Detect fraud in real-time with Azure Cosmos DB, AKS, and Azure AI services.

In this scenario, payment transactions are processed by Azure Cosmos DB, with real-time analytics using Synapse Link/Microsoft Fabric mirroring. Financial transactions are integrated with Microsoft Fabric using Data Factory in Fabric and stored in OneLake. GenAI powers the fraud analysis tool using retrieval augmented generation (RAG), with Azure OpenAI Service for the LLM, and Azure AI Search as the retrieval system, enabling real-time fraud alerts and deeper analysis over client profiles.

Additional resources:

Support bots and information discovery

Service and support bot applications use AI to improve customer service interactions and processes. Such applications can provide personalized experiences to customers and help users extract key information and insights from large volumes of data. Techniques include: natural language processing (NLP) for understanding text and sentiment analysis; machine learning for identifying patterns and making predictions; knowledge mining for exploring and analyzing structured and unstructured data; and RAG for retrieving data/documents relevant to a query or task and providing them as context for the LLM. Example applications include:

  • Chatbots and virtual assistants: Can answer questions, extract knowledge, and help with
  • Intelligent search engines: Can provide more accurate and relevant search results
  • Product catalog discovery: Use of NLP, image analysis, and user preference analysis to provide product recommendations.
  • Drug discovery: AI can be used to optimize drug design and testing, predict drug properties, and enable molecular simulations.
  • Automated report generation and summarization: Can analyze data, extract information, and present it in a structured format.
  • Voice and speech recognition: Enable automated call routing and voice-based interactions.

Let's review some ways that real businesses are employing AI-powered support bots and information discovery.

TomTom, a leader in maps and location technology, used GenAI to create their Digital Cockpit, a conversational automotive assistant enabling voice interaction with infotainment, location search, and vehicle commands. Using Azure OpenAI Service, Azure Cosmos DB, and AKS, they built an intelligent, fast, and highly scalable AI chatbot that can be integrated into other automotive systems.

In another example, Inc., a Japanese company that operates professional services websites, built an AI-powered legal consultation chatbot using Azure Cosmos DB, AKS, and Azure OpenAI Service. When users ask the chatbot a question, it retrieves and summarizes pages in law books that are relevant to the query, helping professionals cut down on legal research time.

Lastly, by removing data silos and going fully cloud native with services such as Azure Cosmos DB, H&R Block has been able to unlock AI-powered information discovery capabilities. Aiming to streamline tax filing and make it less stressful for both clients and tax professionals, H&R Block uses Azure Machine Learning and Azure AI to provide detailed client history in seconds. Accessing this information fast allows H&R Block tax pros to provide a differentiated client experience.

Build your own solution

Below is a sample architecture for a service and support bot in an airline customer service scenario:

Azure architecture for service and support chatbot application.Azure architecture for service and support chatbot application.

This intelligent application has two interfaces – a customer-facing online support portal and an internal Customer Support System application. Azure Front Door provides a single-entry point ( for multi-region architecture. Raw data from flight booking systems, luggage tracking, customer account data, and travel policies are stored in Azure Cosmos DB. AKS hosts the web UI and integrates other components of the solution. The application uses RAG, with Azure AI Search as the retrieval system and Azure OpenAI Service providing Large Language Model (LLM) capabilities, allowing customer service reps and customers to ask questions using natural language.

Additional resources:

Personalization and recommendations

AI-based personalization and recommendation apps tailor content, products, or services to individuals based on their behavior, preferences, and needs. Customers expect top-notch digital experiences today – personalization helps businesses deliver on this by providing real-time recommendations and dynamic customer engagement. Examples of personalization include:

  • Product recommendations in e-commerce/retail: AI engines analyze customer behavior, purchase history, and preferences to provide product recommendations, enhancing the shopping experience.
  • Streaming content recommendations: Platforms like Netflix and Spotify use AI to provide personalized content suggestions, helping users discover new content aligned with their interests.
  • Targeted advertising: AI algorithms analyze user data, browsing behavior, and demographic information to deliver targeted and personalized ads, improving ad relevance and click-through rates.
  • Financial recommendations: AI provides personalized advice on investment, retirement planning, and loan offers, based on individual financial goals, risk profiles, and spending patterns.
  • Personalized learning: Educational platforms use AI to provide personalized learning paths, adaptive assessments, and customized content to students, helping improve educational outcomes.

Sitting on a treasure trove of game and player data, the NBA used Azure to build an AI-powered content personalization engine to elevate the digital fan experience. The NBA CourtOptix platform leverages Azure AI services, Azure Machine Learning, AKS, and Azure Cosmos DB to turn billions of data points into insightful metrics about players. It personalizes these statistical insights into content that is tailored to fan preferences.

ASOS, a global online retailer with more than 26 million active customers, runs a microservices architecture on Azure, with Azure Cosmos DB and Azure Service powering a real-time recommendations engine at enormous scale for ASOS shoppers. The retailer is further enhancing its customer experience by using Azure OpenAI Service and prompt flow in Azure AI Studio to create a natural language-based conversational interface to curate product selections for shoppers.

Build your own solution

Check out this reference architecture for a product recommendation engine.

Azure architecture for real-time recommendations engine.Azure architecture for real-time recommendations engine.

In this scenario, product data, customer profiles, and other information are stored in Azure Cosmos DB, while static content is stored in Azure . Azure AI Search creates a semantic vector index and acts as the retrieval system for Azure Machine Learning. Azure Machine Learning provides machine learning models for product recommendations, creating customer personas and identifying available products and promotions.

Additional resources:

Copilots for everything, everyone

AI copilots are an exciting type of intelligent application that is transforming the way we create and communicate. Copilots go beyond traditional chatbots. They enable rich and contextual interactions and can extract insights from vast amounts of data in an interactive natural language user interface.

With the emergence and massive popularity of ChatGPT, nearly every industry is looking to take advantage of generative AI applications – and copilots are the answer. These large language model-based assistants are unlocking the art of the possible across different domains and for a variety of users by enabling them to perform tasks more efficiently and effectively.

AI copilots can power intelligent use cases across industries.

The opportunities to utilize copilots are endless. They can power nearly every other use case highlighted in this blog and more.

You can build an AI chatbot copilot, a product recommender copilot, a speech and image recognition copilot, a copilot to create and optimize code for a variety of industries, a content generation and summarization copilot, and so on. No less important, copilots can integrate with different applications and platforms to enable seamless communication, data exchange, and enhanced user experiences.

Among the most common examples, organizations are using copilots to:

  • Chat with their data: Interacting with data with natural language to summarize content and get faster access to critical information.
  • Generate content dynamically: Employing AI to optimize and update existing creative, and build new creative assets.
  • Discover information: Interactive user interface can answer common questions, provide information, and help with .
  • Personalize experiences: Offering product and content recommendations to individual users, based on their behavior and preferences.

KPMG Australia developed their own copilot app, KymChat, a conversational AI assistant, to help the firm's 10,000 employees improve productivity. The ChatGPT-based AI assistant helps employees automate mundane tasks and surface insights from the company's external and internal websites, knowledge bases, and Microsoft 365 files. To keep up with the growth of the solution, KPMG Australia used Azure Cosmos DB for MongoDB vCore, Azure OpenAI Service and Azure App Service, enabling KymChat to deliver faster and higher-quality search results at scale.

Build your own solution

Here is an example architecture for an intelligent agent copilot within a consumer retail scenario. The app allows users to ask questions on vectorized product, customer, and sales order data stored in the database.

Build your own copilot/AI assistant with Azure AI, data, and app services.Build your own copilot/AI assistant with Azure AI, data, and app services.

In this scenario, customers of a retail bike store interact with an AI agent to ask questions ranging from product recommendations to information on customer profiles and sales orders. The solution is designed to also enable fast updates to data, demonstrating how adding or updating an item in the product catalog can be used in near real-time by the AI agent.

Additional resources:

Getting started

Microsoft offers AI apps solution accelerators with pre-built templates, code, and to help you implement common AI copilots faster and easier. Each accelerator presents a broadly applicable solution architecture with an easily-understandable, real-world example:

See all solution accelerators and code samples.

Finally, good news: You can save big when building your AI apps on Azure. Check out the offers below to accelerate your AI journey.

  • Azure AI Advantage: Azure AI and GitHub Copilot customers can save up to $6000 on Azure Cosmos DB for 90 days.
  • Azure Innovate: Get expert guidance and funding to build AI-powered and custom cloud-native apps.
  • Try Azure Cosmos DB for free: Develop, test your applications, and run small production workloads for free.

What intelligent apps are you building on Azure? Drop a comment to share your use case.


This article was originally published by Microsoft's Azure AI Services Blog. You can find the original article here.