Proactive Contract Alerts in Teams with Azure AI Search and Speech-to-Text - Build Intelligent Event-Driven Agents

⏱️ 70 minutes 📊 Level 200 🏷️ Local

Configure Copilot Studio agents to proactively notify users, search contract documents, and transcribe audio files using Azure AI services.

🧭 Lab Details

Level	Persona	Duration	Purpose
200	Maker	60–75 minutes	After completing this lab, participants will be able to configure Copilot Studio agents to proactively notify users when new files are added to Azure Blob Storage, connect Azure AI Search as a knowledge source to enable contextual question-answering over documents, integrate Azure Speech-to-Text services to automatically transcribe audio files, and understand how to combine low-code and Azure AI-based services to extend Copilot Studio functionality end-to-end.

📚 Table of Contents

🧭 Lab Details
🎯 Purpose
🤔 Why This Matters
🌐 Introduction
💼 Real-world example
🎓 Core Concepts Overview
📄 Documentation and Additional Training Links
✅ Prerequisites
🎯 Summary of Targets
🧩 Use Cases Covered
🛠️ Instructions by Use Case
🧱 Use Case #1: Event-Driven Contract Notifications
🔄 Use Case #2: Intelligent Document Search and Audio Transcription
🏆 Summary of learnings
⚙️ Azure Setup (For Reference Only)

🎯 Purpose

After completing this lab, participants will be able to:

Configure Copilot Studio agents to proactively notify users when new files are added to Azure Blob Storage.
Connect Azure AI Search as a knowledge source to enable contextual question-answering over documents in Azure Blob Storage.
Integrate Azure Speech-to-Text services to automatically transcribe audio files and return insights through Copilot Studio.
Understand how to combine Low-code and Azure AI-based services to extend Copilot Studio functionality end-to-end.

🤔 Why This Matters

This lab demonstrates how Copilot Studio can move beyond simple chat-based interactions into real-world automation and intelligence.

By connecting to Azure services like Blob Storage, AI Search, and Speech, makers learn to:

Bridge data and conversation — allowing copilots to proactively respond to real business events such as new contracts or uploaded files.
Add enterprise-grade knowledge — enabling copilots to access and reason over indexed data sources securely.
Infuse AI capabilities — such as document understanding and transcription, without custom coding.

Together, these skills showcase how Copilot Studio + Azure AI can transform business workflows into intelligent, event-driven experiences that blend automation, knowledge, and natural language.

🌐 Introduction

As organizations adopt AI copilots across departments, the real value comes not from basic Q&A, but from connecting those copilots to real business data and workflows.

Modern enterprises generate information continuously—from uploaded contracts and customer calls to indexed knowledge in Azure. However, unless your copilot can access, interpret, and act on that data, its usefulness remains limited.

This lab focuses on closing that gap — teaching you how to turn Copilot Studio into a truly data-aware, event-driven assistant powered by Azure AI and Power Platform.

This lab series (Lab 2) is designed to show how Copilot Studio can be extended through real-world Azure AI integrations. Each sub-lab builds on the previous one to help makers understand how to connect conversational agents to enterprise data, events, and intelligence.

The lab is divided into three guided modules (2A, 2B, and 2C) plus an Azure setup reference section:

Lab 2A – Notify users when new contract documents are added to Azure Blob Storage
Participants build an event-driven agent flow that monitors an Azure Blob container and proactively sends Teams notifications when new files are uploaded. This introduces the concept of trigger-based automation within Copilot Studio.
Lab 2B – Configure Azure AI Search on Blob Storage
Attendees connect Copilot Studio to a pre-created Azure AI Search index so their copilot can retrieve and summarize content from contract documents. This demonstrates knowledge integration and semantic search capabilities.
Lab 2C – Speech-to-Text Integration with Copilot Studio
Participants integrate Azure Speech Service with Copilot Studio to transcribe uploaded audio files and return the transcript through the agent. This module adds AI-driven audio understanding to the copilot's skillset.
Azure Setup (Reference)
The appendix provides background on how the Azure resources (Blob Storage, AI Search, OpenAI embedding model, and Speech Service) were provisioned. For this workshop, all Azure components are pre-created, and attendees are supplied with the necessary endpoints and keys so they can focus entirely on building and testing within Copilot Studio.

Together, these modules illustrate how to evolve a Copilot from a simple chat interface into a connected, intelligent assistant that can see, search, and listen across enterprise data sources.

💼 Real-world example

Imagine a sales operations team that handles dozens of new contracts and client calls every week.

Before:

They manually check shared folders for new contracts
Send an email or provide updates in Teams manually
Listen to entire recordings just to confirm details

After: With the capabilities built in this lab, their Copilot automatically:

Detects new contracts uploaded to Azure Blob Storage.
Extracts key details like contract number, vendor, and effective date and posts a short update in Teams.
Answers questions like “What’s the renewal term for Contoso?” using Azure AI Search.
Transcribes customer meeting recordings into text, allowing instant summaries or quick keyword searches.

🎓 Core Concepts Overview

Concept	Why it matters
Event-driven Copilot Flows	Enables copilots to react automatically to real-world triggers (like new files in Azure Blob Storage). This bridges the gap between conversational AI and operational workflows, delivering faster user notifications and reducing manual tracking while maintaining enterprise security and compliance.
Connecting Azure Services to Copilot Studio	Shows how to extend copilots beyond chat by connecting them securely to Azure resources (Blob Storage, AI Search, Speech). This integration gives copilots access to enterprise data and intelligence while maintaining governance, enabling true digital transformation of business processes.
Azure AI Search as Knowledge Source	Demonstrates how to add contextual understanding by letting copilots query and summarize enterprise documents. This makes business data instantly searchable and conversational, improving productivity and decision-making by transforming static documents into interactive knowledge.
Speech-to-Text Integration	Introduces real-time AI transcription capabilities so copilots can understand and summarize customer calls or meetings. This adds "hearing" to your Copilot—turning unstructured audio into structured insights for teams, enabling faster follow-ups and better customer service.

📄 Documentation and Additional Training Links

✅ Prerequisites

Access to Microsoft Copilot Studio, with permissions to create, edit, and publish agents and agent flows.
Contoso Agent created in Lab 1 (or equivalent Copilot Studio agent).
Microsoft Teams account to test Copilot notifications and interactions.
Provided connection details (endpoint URLs, access keys, and container names) for pre-created Azure resources—including Blob Storage, AI Search, and Speech Services.
Power Platform environment enabled for Copilot Studio.
Basic familiarity with Copilot Studio concepts such as agent setup, knowledge sources, and flows.

[!NOTE] All Azure related resources are pre-provided for this lab. However we do provide instructions in this section Azure Setup (For Reference Only) section in the Appendix, if you want to set up these resources in your own subscription.

🎯 Summary of Targets

In this lab, you'll transform your Copilot Studio agent into an intelligent, event-driven assistant that connects to Azure AI services. By the end of the lab, you will:

Configure proactive notifications triggered by new files in Azure Blob Storage with AI-powered content extraction.
Connect Azure AI Search as a knowledge source to enable semantic search over contract documents.
Integrate Azure Speech-to-Text services for automatic audio transcription and insights.
Understand the architecture of combining low-code Power Platform tools with Azure AI services.
Apply event-driven automation patterns that extend beyond simple chat interactions.

🧩 Use Cases Covered

Step	Use Case	Value added	Effort
1	Event-Driven Contract Notifications	Automate contract processing with proactive Teams notifications and AI-powered content extraction	35 min
2	Intelligent Document Search and Audio Transcription	Enable semantic document search and speech-to-text capabilities for comprehensive business intelligence	30 min

🛠️ Instructions by Use Case

🧱 Use Case #1: Event-Driven Contract Notifications

Build an intelligent agent flow that automatically detects new contract uploads and sends proactive Teams notifications with AI-extracted key details.

Use case	Value added	Estimated effort
Event-Driven Contract Notifications	Automate contract processing with proactive Teams notifications and AI-powered content extraction	35 minutes

Summary of tasks

In this section, you'll learn how to create event-driven agent flows that monitor Azure Blob Storage, extract contract information using AI Builder, and send proactive notifications to Teams.

Scenario: A sales operations team needs automatic notifications when new contracts are uploaded, with key details extracted and shared in Teams without manual intervention.

Objective

Create and configure an event-driven flow that monitors Azure Blob Storage for new contracts, extracts key information using AI, and sends proactive Teams notifications.

Step-by-step instructions

Lab 2A: Notify Users When New Contract Documents Are Added to Azure Blob Storage

In this lab, you’ll create an event-driven Agent Flow that automatically alerts users in Microsoft Teams whenever a new contract document is uploaded to Azure Blob Storage. Using Agent Flows within Copilot Studio, you’ll connect your agent to Blob Storage, extract key contract details using custom prompts, and send a neatly formatted Teams message. By the end of this module, your copilot will be able to detect new files, understand what they contain, and proactively notify users — bringing real-time awareness into your business workflows.

Prepare and Publish Your Agent

2A.1 Open the Contoso Agent that was built in Lab 1 and Publish it.

[!IMPORTANT] We are required to publish the Agent to Teams and Microsoft 365 Copilot channel so that the agent can message/ping proactively in Teams.

Publish button in Copilot Studio agent overview page

2A.2 Select the Channels tab and click on Teams and Microsoft 365 Copilot, then select Add channel.

Channels tab showing Teams and Microsoft 365 Copilot option with Add channel button

Create Event-Driven Agent Flow

2A.3 Select Flows and click on + New agent flow to add a new Agent Flow to notify users when new documents are uploaded to blob storage.

Flows tab with New agent flow button highlighted

2A.4 In the designer, select Add a trigger node and search for blob. Select When a blob is added or modified (properties only) (V2) trigger.

Flow designer showing Add a trigger node with blob search results and trigger selection

2A.5 Select change connection reference to add a new connection. Provide the following details to connect to Azure Blob Storage account:

Connection name: ContractsBlobStorage
Authentication type: Access Key
Azure Storage Account Name: [Provided in Lab Resources]
Azure Storage Account Access Key: [Provided in Lab Resources]

Azure Blob Storage connection configuration dialog with connection name, authentication type, and account credentials

2A.6 Once the connection is setup, use the dropdown to select the Storage account name in the container.

Blob trigger configuration showing storage account name dropdown selection

Configure Blob Content Processing

2A.7 Add a new action – Get Blob Content (V2) in the Azure Blob Storage actions to the flow.

Flow designer showing Add an action with Get Blob Content V2 action selected

2A.8 Since the connection is already established to blob storage account, you can select the Storage account name from the dropdown menu. For the Blob field, type / and then container name. Then / and add the Dynamic content from the trigger node – body/Name (Ex: /{Container Name}/{Dynamic Content}).

Get Blob Content configuration with storage account dropdown and Blob field showing dynamic content path

2A.9 We will be using pre-built AI builder prompt – Extract information from Contract to extract all the information from the documents. Add a new action in the flow and select action - Extract information from Contract.

Flow designer showing Add an action with Extract information from Contract AI Builder action

2A.10 In the Extract information from Contract action parameters, select Dynamic content as File Content (from Get Blob content node) into Contract file field.

Extract information from Contract action with dynamic content selector showing File Content being mapped to Contract file field

Create Custom AI Prompt for Notifications

2A.11 We will use another custom AI builder prompt to extract the key details (ContractID, Customer Name, Vendor Name and Effective Date) for the notification. Add a new action in the flow and select action – Run a prompt.

Flow designer showing Add an action with Run a prompt AI Builder action

2A.12 Click the dropdown in the Prompt and then +New custom prompt under Prompt parameter to create a new prompt.

Run a prompt action configuration with Prompt dropdown expanded showing New custom prompt option

2A.13 In the new pop-up prompt window, update the prompt name to – Extract Contract Info.

Copy and paste the following in the instructions:

Extract Contract Number, Customer name, Vendor name and Date from {ContractInput} and Provide extracted information like following:  

New Contract Available for Review:
Contract Number: AW2024-003
Customer Name: Adventure Works
Vendor Name: Contoso Solutions Inc.
Date (Effective Date): March 10, 2024

Custom prompt builder showing Extract Contract Info prompt name and instructions with ContractInput parameter

2A.13a Replace {ContractInput} with a new input parameter. Delete the text in the instructions and keep your cursor in that spot. Click + Add Content, select Text and name it ContractInput. Click Close. Your instructions should now look the screenshot above. Click Save.

Add Content dialog showing Text input type selection for creating ContractInput parameter

2A.14 Select the newly created prompt in the dropdown for Prompt and pass the dynamic content Body (from Extract information from contract node) into ContractInput.

Run a prompt action showing Extract Contract Info prompt selected with Body dynamic content mapped to ContractInput field

[!TIP] If you do not see the Body from the Extract information from contract node action, you can click See more to view all the dynamic content.

Dynamic content panel expanded showing See more link to reveal additional content options including Body

[!TIP] If you still don't see the Body, click on Insert Expression and paste the following expression:
body('Extract_information_from_Contract')
and click Add.

Expression editor dialog with body expression for Extract_information_from_Contract action

Send Teams Notification

2A.15 Add a new action into the flow – Post a message in a chat or channel to send a message to a user on Teams regarding the new contract document. Configure the parameters as:

Post as: Power Virtual Agents (preview)
Post in: Chat with bot
Bot: Contoso Agent
Recipient: <Your lab user account>
Message: Text (Insert Dynamic Content and Select from Run a prompt action)

Post a message in a chat or channel action configured with Power Virtual Agents, Contoso Agent bot, recipient, and dynamic message content

2A.16 Save Draft and Switch to the overview tab and click edit to change the name of the flow to – New Contracts Notification and select Save.

Flow overview tab showing edit button for renaming the flow

Flow name editor showing New Contracts Notification as the flow name with Save button

2A.17 Switch to the Designer tab and Click on Publish to Publish the Agent flow.

Publish button in agent flow interface to deploy the flow

Testing Lab 2A

As new documents are added to Blob Storage, you should see a proactive notification from your Contoso Agent with key details as shown below. Open Teams in the browser using your lab credentials and open your Contoso Agent.
The proctors will be adding documents to the Azure Blob storage every 5 – 10 minutes to trigger the agent flow.

Teams chat showing proactive notification from Contoso Agent with extracted contract details including contract number, customer name, vendor, and effective date

Try asking for a follow-up question on this contract and the note that the agent does not provide a good answer yet!
This is what we will achieve next in Lab 2B.

Teams chat showing follow-up question about contract with agent unable to provide detailed answer without knowledge source

🏅 Congratulations! You've completed Event-Driven Contract Notifications

Test your understanding

Key takeaways:

Event-Driven Automation – Blob storage triggers enable real-time responses to business events, eliminating manual monitoring and notification processes.
AI-Powered Content Extraction – AI Builder prompts can automatically extract structured information from unstructured documents, making data immediately actionable.
Proactive Communication – Copilot Studio agents can initiate conversations in Teams, bringing intelligence directly into collaborative workspaces.

Lessons learned & troubleshooting tips:

Agents must be published to Teams channel before they can send proactive messages
Use dynamic content carefully and check "See more" if expected options don't appear
AI Builder prompts require clear instructions and proper input parameter mapping

Challenge: Apply this to your own use case

What other document types in your organization could benefit from automated processing and notification?
How would you modify the AI prompt to extract different types of business-critical information?
Consider what other Azure services could trigger similar automated workflows.

🔄 Use Case #2: Intelligent Document Search and Audio Transcription

Connect Azure AI Search for semantic document search and integrate Speech-to-Text services for comprehensive business intelligence capabilities.

Use case	Value added	Estimated effort
Intelligent Document Search and Audio Transcription	Enable semantic document search and speech-to-text capabilities for comprehensive business intelligence	30 minutes

Summary of tasks

In this section, you'll learn how to connect Azure AI Search as a knowledge source for semantic document search and integrate Azure Speech Services for automatic audio transcription.

Scenario: Users need to ask natural language questions about contracts and get instant answers, plus the ability to transcribe customer call recordings for quick insights and searchable content.

Step-by-step instructions Lab 2B

Lab 2B: Configure Azure AI Search on Blob Storage

In this lab, you’ll connect your Copilot Studio agent to Azure AI Search so it can intelligently retrieve and summarize information from contract documents stored in Azure Blob Storage.
You’ll use a pre-created search index and link it as a knowledge source for your agent, giving it the ability to answer natural language queries like What are the renewal terms for Fourth Coffee’s contract?
By the end of this module, your copilot will be able to search, understand, and respond using real contract data—transforming static documents into conversational knowledge.

Pre-requisite: Azure AI Search configured in Azure portal with text embedding model on Azure Blob storage account where all contract documents are stored

2B.1 Open Contoso Agent and click on +Add Knowledge on the overview tab.

Copilot Studio agent overview page with Add Knowledge button highlighted

2B.2 Select Azure AI Search.

Knowledge source selection dialog showing Azure AI Search option

2B.3 Click on Your connections and select Create new connection.

Azure AI Search connection dialog with Your connections tab and Create new connection option

2B.4 Use the provided Endpoint URL and Access Key to connect to Azure AI Search service:

Endpoint URL: https://ppcaisearch001.search.windows.net
Azure AI Search Admin Key: [[Provided in Lab Resources]

Azure AI Search connection configuration with Endpoint URL and Admin Key fields

2B.5 Select the available index and select Add to agent.

Azure AI Search index selection showing available indexes with Add to agent button

2B.6 Once the Azure AI Search service is connected as knowledge source, select it to edit the Name and Description.

Knowledge sources list showing connected Azure AI Search with edit option

2B.7 Update the following and then click Save:

Name: Contoso Contracts

Description:

   This knowledge source provides access to customer contract agreements. Each 
   contract file follows the format: Customer Contract Agreement - [Customer Name].docx 
     
   Each contract contains essential details, including: 
   • Scope of Services: Defines the services provided  
   • Payment Terms: Specifies the total contract value, milestone-based payments, and due dates. 
   • Service Level Agreement (SLA): Outlines uptime guarantees, response times, and escalation procedures. 
   • Data Privacy & Compliance: Ensures compliance with regulations like GDPR, CCPA, and defines data ownership. 
   • Contract Term & Renewal: States the contract duration, renewal policies, and termination conditions. 
   • Governing Law: Specifies applicable legal jurisdiction

Knowledge source configuration editor showing Name as Contoso Contracts and detailed Description of contract content

2B.8 Go to Agent Overview tab and click Edit to add additional instruction to the agent instructions, then click Save:

   For Contract related queries, use the Contoso Contracts knowledge source

Agent overview showing Instructions editor with added instruction for contract queries using Contoso Contracts knowledge source

Testing Lab 2B

In your test window, send a message like: Contract terms and renewal details for Fourth Coffee

Teams test chat showing user query about Fourth Coffee contract with agent response providing contract terms and renewal details from Azure AI Search knowledge source

Lab 2C: Speech-to-Text Integration with Copilot Studio

In this lab, you’ll integrate your copilot with Azure Speech Services to transcribe customer or meeting recordings directly within Copilot Studio.
You’ll build a flow that uploads an audio file, sends it to Azure Speech for transcription, retrieves the text, and returns it to the user through the agent.
By the end of this module, your copilot will be able to listen and understand audio content, turning long recordings into instantly searchable text and actionable insights.

Pre-requisites:

Azure Speech Service created
Azure Blob container to hold audio files

2C.1a In Copilot Studio, open Agent Flows, Select +New agent flow.

Agent Flows page with New agent flow button highlighted

2C.2 Add a Trigger – When an agent calls the flow.

Flow trigger selection showing When an agent calls the flow trigger option

2C.3 Add a new action Create transcription (V3.1) from Azure Batch Speech-to-text actions.

Flow designer showing Add an action with Create transcription V3.1 from Azure Batch Speech-to-text connector

2C.4 Create connection with provided details then click Create new:

Connection name: SpeechtoTextConnection
Auth. Type: API Key
Account Key: [[Provided in Lab Resources]
Region: eastus

Azure Speech Service connection configuration with connection name, API Key authentication, account key, and region fields

2C.5 Add the following Action parameters for Create transcription (V3.1):

Transcription/locale: en-US
Transcription/displayName: Call transcription
transcription/contentUrls: (Advanced Parameters) [[Provided in Lab Resources - Audio File URL]

Create transcription action configuration showing locale, displayName, and contentUrls parameters

2C.6 Add an Initialize variable action with Parameters:

Name: TranscriptionID
Type: String
Value: trim(last(split(outputs('Create_transcription_(V3.1)')?['body/self'], '/'))) (Insert expression)

Initialize variable action configured with TranscriptionID name, String type, and expression to extract transcription ID

2C.7 Add a Delay action with Count = 30 and Unit = Second.

Delay action configured with 30 seconds count and Second unit to allow transcription processing time

2C.8 Add a Get transcriptions list files (V3.1) from Azure Batch Speech-to-text actions and pass the TranscriptionID variable as Dynamic Content from the earlier step.

Get transcriptions list files action with TranscriptionID variable mapped from previous Initialize variable step

2C.9 Add a Get transcription file (V3.1) action with Parameters:

Id: variables('TranscriptionID')
File Id: last(split(first(body('Get_transcriptions_list_files_(V3.1)')?['values'])?['self'], '/files/')) - Add this using the Insert expression

Get transcription file action with Id using TranscriptionID variable and File Id expression to extract file identifier

2C.10 Add an HTTP action with:

URI: body('Get_transcription_file_(V3.1)')?['links']?['contentUrl']

[!NOTE] You will not see the icon from the Get transcription file until it resolves. It will first show as an expression.

Method: GET
Headers Key: Accept
Headers Value: application/json

HTTP action configured with URI from Get transcription file contentUrl, GET method, and Accept application/json header

2C.11 Add a Compose action. In the Inputs field paste: body('HTTP')?['combinedRecognizedPhrases']

[!NOTE] You will not see the icon from the HTTP action until it resolves. It will first show as an expression.

Compose action with Inputs field containing expression to extract combinedRecognizedPhrases from HTTP response body

2C.12 Add an Action - Respond to the Agent with a Text output parameter: - Name: Transcript - Value: first(outputs('Compose'))?['display']

Respond to the Agent action configured with Transcript output parameter using expression to extract display text from Compose output

2C.13 Click on Save draft to save the agent flow. Go to Overview tab and update the flow name to – Transcribe Customer Call and Publish the flow.

Flow overview showing Save draft button, name editor with Transcribe Customer Call, and Publish button

2C.14 Open the Contoso Agent, go to Tools tab and select +Add a tool.

Copilot Studio agent Tools tab with Add a tool button highlighted

2C.15 In the Add tool window, Select Flow filter and select the newly created Transcribe Customer Call flow and add to the agent by selecting Add to agent button.

Add tool dialog with Flow filter selected showing Transcribe Customer Call flow and Add to agent button

2C.16 You should see the flow added as shown below

Tools tab showing successfully added Transcribe Customer Call flow in the agent's tool list

2C.17 In the Contoso Agent Overview tab, update instructions to direct the agent to this Agent flow for customer call insights as shown below. .

Agent instructions editor showing how to reference the Transcribe Customer Call flow for call insights

To get the actual tool listed in the instructions. Type / and then scroll down and select the Transcribe Agent Flow.

Instructions editor showing slash command to insert Transcribe Customer Call flow reference into agent instructions

Testing Lab 2C

In your Agent Test pane, send a message:

Customer call insights.

(If prompted, create the required connection and retry)

Agent test pane showing user query for customer call insights with transcription response containing full audio transcription text

🏅 Congratulations! You've completed Intelligent Document Search and Audio Transcription

Test your understanding for Azure AI and Speech to Text

How does Azure AI Search improve the agent's ability to answer contract-related questions compared to basic document storage?
What are the key components required for Speech-to-Text integration, and how do they work together?
How would you extend this approach to handle different audio formats or languages?

Challenge: Apply this to your own use case

Consider what other document types in your organization could benefit from semantic search capabilities
Think about different audio sources (meetings, voicemails, interviews) that could be transcribed and analyzed
Explore how you could combine search results with transcription insights for comprehensive business intelligence

🏆 Summary of learnings

True learning comes from doing, questioning, and reflecting—so let's put your skills to the test.

To maximize the impact of intelligent, event-driven Copilot Studio agents:

Event-Driven Architecture – Design agents that respond automatically to real business events, eliminating manual monitoring and creating proactive workflows
AI-Powered Content Processing – Leverage AI Builder and Azure AI services to extract structured insights from unstructured data sources
Multi-Modal Intelligence – Combine document search, speech transcription, and conversational AI to create comprehensive business intelligence solutions
Secure Azure Integration – Connect Copilot Studio to enterprise Azure services while maintaining security, compliance, and governance standards
User-Centric Design – Build agents that deliver intelligence directly in collaborative workspaces like Teams where users already work

Conclusions and recommendations

Intelligent Agent Development golden rules:

Always publish agents to appropriate channels before implementing proactive messaging capabilities
Design AI prompts with clear instructions and structured output formats for consistent results
Implement proper error handling and delays when integrating with external Azure services
Use semantic search and knowledge sources to ground agent responses in enterprise data
Test integration points thoroughly with realistic data and scenarios
Consider the full user workflow when designing proactive notifications and responses

By following these principles, you'll create powerful, intelligent agents that seamlessly integrate enterprise data, events, and AI capabilities into natural conversational workflows—transforming how organizations process information and respond to business events.

⚙️ Azure Setup (For Reference Only)

This section provides background on the Azure resources that power your agent capabilities—Blob Storage, Azure AI Search, Azure OpenAI embeddings, and Speech Services. You don't need to perform these steps for the lab as these resources are already pre-created for this lab. This reference helps you understand how each component fits into the overall solution architecture. You'll see how contract documents are stored, indexed, and transcribed behind the scenes—so you can replicate or extend this setup in your own environment later.

A. Create Azure Blob Storage Account with Container

Prerequisites:

Azure Subscription

Step-by-step instructions:

Login to Azure Portal, in the home screen, click on Create a resource and search for storage account
Create a storage account by selecting the right values
Once the storage account is created and ready, click on Go to resources
Under data storage, select Containers and select + Add container
Add a name for the container and click Create
In the storage account resource, select Access keys under security + networking tab
Copy the storage account name and one of the key values for connecting to it from the Agent Flow

B. Azure AI Search on Contracts Stored in Azure Blob Storage

Pre-requisite: Ensure all the contract documents are added to the container inside the storage account

Azure Blob Storage container showing uploaded contract documents

In the following steps, we will be creating:

Create Azure OpenAI resource and deploy an embedding model
Create Azure AI Search resource
Import & vectorize Blob data with Azure AI Search
Obtain endpoint and keys for the Search service

Step-by-step instructions:

Go to the Azure Portal home screen and click on Create a resource
Search for Azure OpenAI and select it to create a new Azure OpenAI resource
Click Create
Select right values for creating an Azure OpenAI resource and click on Next until the last step
Review values and select Create
Once the Azure OpenAI resource is ready, click on Go to resource
Select Explore Azure AI Foundry portal (This should open a new browser window)
In the Azure AI Foundry portal, ensure the right Azure OpenAI resource is selected. Click on Deployments
Select text-embedding-ada-002 model for deployment and confirm
Ensure the deployment details reflect the right values and click on Deploy
Go to the Azure portal home page and select Create a resource
Search for search service and select Azure AI Search
Click on Create
Configure search service with right values and select Review + Create
Review the values and select create
Once the search service is deployed, select Go to resource
Select Import data (new) option
Select Azure blob storage as data source as we're configuring the search on documents in the blob
Select RAG scenario
Configure values to connect to the data source (Blob storage) and select Next
Select values to vectorize your text and select Next until the last step
Update the index name, review all details and click Create
Once the data is connected to search service, select the Overview tab in your search service and copy the URL (We will use this in Copilot Studio for connecting to search service)
Expand settings tab and click on Keys, copy one of the key values (We will use this in Copilot Studio for connecting to search service)

C. Create Azure Speech Service and Azure Blob Container for Audio Files

In Azure portal, search and select Azure AI services resource
Select Create
Configure Azure AI service as shown below and create it
Once the resource is created, click on Go to resource
Select keys and Endpoint under Resource management to copy the Key and Location/Region values
Select the existing Blob Storage account and add a new container to host audio files
Provide a container name and click create
Upload an audio file (ex: Customer call recording) into the container
Select the uploaded audio file and Click on Generate SAS to get the Blob SAS URL