Skip to main content
Deploy AI for Work directly inside your organization’s collaboration platforms and host applications. This guide covers integration with Microsoft Teams, Slack, and the Web/Mobile SDK, including architecture, prerequisites, and step-by-step configuration.

Microsoft Teams

AI for Work integrates with Microsoft Teams using Azure Bot Service, enabling users to access AI assistance, receive proactive notifications, and use Copilot features without leaving Teams.

When to Use

Architecture

The integration uses Azure Bot Service as the communication bridge:

Prerequisites

Configuration

Follow these phases sequentially to ensure proper deployment and functionality. Phase 1: Initial Setup and Webhook Configuration Objective: Establish communication between AI for Work and Azure Bot Service.
  1. Log in to the Admin Console.
  2. Navigate to Channels > Microsoft Teams and Copilot.
  3. Copy and save the Webhook URL, to use in Azure Bot configuration later.
Phase 2: Azure Bot Service Creation Create and configure the Azure Bot that serves as the communication bridge. Create Azure Bot Resource
  1. Sign in to the Azure Portal.
  2. Click + Create a resource → Search for Azure Bot.
  3. Select Azure Bot and configure the following:
    • Bot Handle: Choose a unique name.
    • Subscription: Select your active subscription.
    • Resource Group: Create or reuse an existing group.
    • Pricing Tier: Select based on expected usage.
    • App Type: Select Multi-Tenant or Single-Tenant based on organizational needs.
Configure Bot Resource
  1. After deployment completes, click Go to resource and navigate to Configuration in the left sidebar.
  2. Paste the webhook URL copied from the AI for Work platform into the Messaging endpoint field and click Apply to save the configuration.
Phase 3: Authentication and Security Setup This phase varies based on your chosen App Type (MultiTenant vs SingleTenant). For MultiTenant Applications Generate Client Secret
  1. In Configuration, click Manage Password next to the Microsoft App ID. teams_1
  2. Go to Certificates & secrets → + New client secret. teams_2
  3. Enter description and expiry → Add.
  4. Copy the Value of the client secret (App Password).
  5. Record the Microsoft App ID and App Password for the Platform configuration.
For Single-Tenant Applications Follow the steps for Multi-Tenant setup. Also, note the Tenant ID for the Platform configuration. Phase 4: Enable Microsoft Teams Channel Integration Configure the bot to communicate with Microsoft Teams and enable Copilot integration.
  1. Navigate to Channels in the bot resource menu and locate the Available Channels section.
  2. Select the Microsoft Teams channel and M365 Extensions channel (required for Copilot integration), review channel information, and select Save. teams_3
This integration doesn’t support the Web Chat channel, so keep it disabled.
Phase 5: Proactive Notifications Setup Enable the bot to send proactive notifications to users, which is essential for AI-driven interactions. Create App Registration for Notifications
  1. In the Azure Portal, navigate to App registrations and select + New registration.
  2. Register New Application:
    • Name: Provide a descriptive name (for example, Platform-Microsoft Teams-Notifications).
    • Supported account types: Choose based on organizational requirements.
    • Redirect URI: Leave blank for this use case.
  3. Select Register. teams_4
Add Microsoft Graph Permissions
  1. In the app registration, navigate to API permissions and select Add a permission. teams_5
  2. Select Microsoft Graph → Application permissions. Required Permissions: Add the following permissions for full functionality:
    • TeamsAppInstallation.ReadForUser.All
    • TeamsAppInstallation.ReadWriteSelfForUser.All
    • TeamsAppInstallation.ReadWriteForUser.All
    • User.Read.All
    • AppCatalog.Read.All
  3. Grant Admin Consent.
  4. Generate a Client Secret and note down App ID, Directory (Tenant) ID.
  5. Configure these values in AI for Work under Microsoft Teams & Copilot Channel Settings.
  6. Generate a JWT token for secure communication.
Phase 6: Configure AI for Work Platform Provide Azure Bot credentials to the Platform to establish secure communication.
  1. Navigate to the Admin Console.
  2. Locate Channels on the left pane.
  3. Select Microsoft Teams and Copilot.
  4. Enter the following credentials under Configuration:
    • Microsoft App Tenant ID: (SingleTenant only) The tenant ID from the bot’s app registration.
    • Microsoft App ID: The App ID from your Azure Bot resource.
    • App Password: The client secret from the bot’s app registration.
    teams_6
    • App ID of the Microsoft Teams App: The additional app credentials for sending Proactive Notifications.
    • Application (Client) ID: From the notifications app registration.
    • Client Secret: From the notifications app registration.
    • Delivery (Tenant) ID: From the notifications app registration.
    teams_7
Customization Options By default, the Microsoft Teams app appears with the name AI for Work, the standard logo, and the default description. Administrators can customize the application before publishing to Microsoft Teams:
  • App Name – Choose a custom display name for Microsoft Teams.
  • Logo – Upload an organization-specific logo.
  • Description – Provide a tailored description that reflects organizational context.
Copilot Enablement During configuration, you are prompted to Enable Copilot Integration. We recommended you to enable this option so that Teams users can access the Platform features using Copilot. Phase 7: Upload Custom App to Teams Admin Center Upload the Platform app package to the Teams Admin Center so it’s available within your organization. Generate Application Package Upon configuring credentials in the Platform, generate and download the Teams application package (ZIP file). This package contains the app manifest and all required configurations for the integration. teams_8 Upload to Teams Admin Center Ensure you have the necessary admin permissions in the Teams Admin Center. Verify organizational policies allow custom app uploads.
  1. Sign in to Microsoft Teams Admin Center.
  2. Navigate to Teams apps → Manage apps.
  3. Click Upload new app and select the downloaded application package. Wait for upload and verify to complete. teams_9
Once uploaded, the app appears as a custom app within your organization’s Teams Admin Center. It isn’t published to the public Teams App Store. Configure App Permissions and Policies Review the uploaded application details and configure app permission policies as needed. Set up app setup policies for automatic installation (optional). Phase 8: User Access Configuration You can decide how to distribute the custom app within your organization:
  • Add App for All Users (Recommended) — Automatically installs the custom app for all employees in the organization. All users receive a notification in Microsoft Teams and a welcome message when the app is added.
Users must also be provisioned in the Account Hub of the Platform to access the app.
  • Let Users Add the App Themselves — The custom app is available in the organization’s app catalog within Teams. Users need to search for the app in the Apps section in Teams and add it manually. No automatic notifications or installations occur.
Users must be present in the Account Hub to use the app.
  • Add App for Selected Users or User Groups — Uses app setup policies to assign the app only to specific users or security groups. Selected users are automatically notified in Microsoft Teams and receive the welcome message. Other users can find and install the app from the organization’s app catalog if permitted by policy.
Selected users must also exist in the Account Hub for access.

Rich Templates

Agent Platform agents can return rich templates, such as lists, tables, charts, forms, carousels, and media. The app renders these templates in Microsoft Teams and Copilot. For the templates that agents can return, see Rich content.

Troubleshooting

Bot Not Responding in Microsoft Teams
  • Verify the webhook URL is correctly configured in Azure Bot.
  • Check that the Microsoft Teams channel is enabled and saved.
  • Validate client secrets haven’t expired.
Authentication Failures
  • Confirm admin consent is granted for all required API permissions.
  • Verify client secrets are copied correctly (Value, not Secret ID).
  • Check tenant ID matches between configurations.
App Upload Failures
  • Ensure you have proper admin permissions in the Teams Admin Center.
  • Verify the application package isn’t corrupted during download.
  • Check organizational policies allow custom app uploads.
Proactive Notifications Not Working
  • Verify that admin consent is granted for all required Microsoft Graph permissions.
  • Verify that the notification app registration client secret is valid.
  • Ensure JWT token generation is working correctly.

Slack

AI for Work integrates with Slack as a custom app, enabling users to perform work tasks, query enterprise knowledge, and interact with AI agents in direct messages and channels without leaving Slack. Slack channel overview

When to Use

Architecture

The integration uses Slack’s Developer Portal and OAuth authentication:

Prerequisites

Integration Lifecycle

Configuration

Step 1: Configure App Display Settings
  • Log in to the Admin Console.
  • Navigate to Channels > Slack. Slack channel navigation
  • Open the Configuration page and set the following: App display settings
If you change the app name or description after deployment, generate a new manifest and update the Slack app in the Developer Portal.
Step 2: Generate the App Manifest
  1. On the Configuration page, click Get Manifest.
  2. Copy the entire manifest code. Manifest generation
The manifest contains all permissions, features, and settings that the Slack app requires. It sets up the app with the Slack agent experience. To move an app that you created with an earlier manifest, see Migrate an Existing App.
Upload the app icon directly to the Slack Developer Portal after you create the app.
Step 3: Create the App in the Slack Developer Portal
  1. Go to the Slack Developer Portal and sign in with your workspace credentials.
  2. On the Your Apps page, click Create New App.
  3. In the Create new app dialog, under Or start your own way, select From a manifest, and then click Continue. Create new app
  4. In the Create from a manifest dialog, select the JSON tab and replace the sample code with the manifest that you copied from the Admin Console.
  5. Under Workspace, select the target workspace, and then click Next. You cannot change the workspace after you create the app. Create from a manifest
  6. In the Review your app dialog, review how the app works and the configurations under What it can do in Slack and What it can respond to in Slack, and then click Create and Install. Review your app
  7. When Slack confirms that the app is ready, click Go to App Settings to open the app in the Developer Portal. App ready confirmation
The confirmation dialog also lists Slack CLI steps for running the app locally. AI for Work does not need these steps.
The app deploys to the selected workspace by default. To deploy to additional workspaces, activate public distribution after you publish the app (see Step 5).
Step 4: Customize the App Icon
  1. In the Slack Developer Portal, navigate to Basic Information.
  2. Scroll to the Display Information section.
  3. Configure the following settings:
    • App Icon: Click Upload and select an image file that represents your organization.
    • Background Color: Customize the background color for the app icon.
    • Short Description and Long Description: Update as needed to refine how the app appears in listings.
  4. Click Save Changes. App icon customization
The app icon appears in all Slack interactions, including the app directory, direct messages, and channel mentions. Step 5: Activate Public Distribution (Optional) Enable public distribution if you manage multiple Slack workspaces and want to deploy the app across all of them.
  1. In the Slack Developer Portal, navigate to Settings > Manage Distribution.
  2. Review the distribution requirements on the screen.
  3. Turn on Public Distribution for the app.
  4. Click Activate Public Distribution.
After you activate public distribution, you can add the app to additional workspaces beyond the one you selected during app creation. Skip this step if you need the app in only one workspace. Step 6: Configure App Credentials Retrieve credentials from Slack
  1. In the Slack Developer Portal, navigate to Basic Information in the left sidebar.
  2. Scroll to App Credentials and copy the following values:
    • Client ID
    • Client Secret
    • Signing Secret
    App Credentials
Enter credentials in AI for Work
  1. Return to the Admin Console > App Credentials page.
  2. Paste the copied values into the corresponding fields.
  3. Click Authorize. Credentials configuration
Complete authorization
  1. Review the permission request that Slack displays.
  2. Click Allow.
After successful authorization, the app deploys to your Slack workspace and all onboarded users can see it. Step 7: Customize the Welcome Experience
  • In the Admin Console, navigate to the Welcome Experience page. Welcome Experience
  • Configure the following:
  • For each prompt, set:
    • Display Message: Text that appears to users as the prompt button.
    • Actual Query: The complete query the Platform receives when the user clicks the prompt.
  • Click Save and Update. Changes take effect immediately for all users.
You can update the welcome experience at any time. Changes apply to all users immediately.
Step 8: Configure Publish Settings Configure how the app behaves in different Slack contexts for each workspace. These settings determine which agents handle conversations and how users can customize channel behavior. Configure Direct Message Behavior
  1. In the Publish Settings page, select the workspace you want to configure.
  2. Under 1:1 App Chat, select the agent that the app uses in direct messages.
Configure Channel Governance Define which users can customize the app behavior in Slack channels.
  1. Under Channel Governance, select the users who can customize Slack channels.
  2. These users can add specific agents to channels and configure how the app responds in those channels.
Set Default Channel Behavior Choose whether the app replies publicly or privately when someone adds it to a channel.
  • Under Default Channel Behavior, select one of the following options:
  • Click Save to apply the default behavior to all channels.
Users with channel customization permissions can override this default for specific channels they manage.
Agent Selection In both Publish Settings and Channel Settings, the Select an Agent list offers the following options: The selected agent does not need access granted to everyone in the account. Channels that used the Work option keep working with no changes and now show Active Enterprise Knowledge Source. If the selected agent doesn’t have access granted to everyone, a note appears below the customization section: “Note: This agent isn’t published or not accessible to everyone in the account.” If a user who can’t access the configured agent sends a query in that channel, that user sees the following message: “The sources of this channel have been customized. You don’t have access to this customization. Please contact @<UserName who added it>” User Access Verification The system displays an alert if some Slack members haven’t onboarded to AI for Work. Add these users from Admin Console > Account Hub > Users before they interact with the app.

Channel Customization

Configure channel-level behavior from the Channel Settings step of the Slack app in the Admin Console. The page has three sections: Customize the agent for a channel
  1. In the Admin Console, open the Slack app and go to Channel Settings.
  2. In the Agent Customisation section, add a configuration and select the channel.
  3. Choose the Reply Type (Private Reply or Public Reply) for how the app responds in the channel.
  4. Select an Agent for the channel:
    • For Private Reply: Select from Search Agents and Autonomous Agents.
    • For Public Reply: Select from Active Enterprise Knowledge Source, Autonomous Agents, and Search Agents.
  5. For Search Agents, optionally scope the Search Sources that the agent uses in that channel.
  6. Click Save to apply the customization to all users in that channel.
Each configuration appears as a compact entry in the Agent Customisation list, in the same layout as Active Listening configurations.

Active Listening

Active Listening lets the app answer questions in a channel without anyone tagging it. The app reads new messages in the channel, decides which ones it can answer, and replies in the reply mode configured for the channel. Active Listening works in channels that use Private Reply or Public Reply. Turn on Active Listening for a channel
  1. In the Admin Console, open the Slack app and go to Channel Settings.
  2. In the Active Listening section, add a configuration and select the channel.
  3. Add instructions that describe which queries the app answers in that channel.
  4. Select Customise for the channel, and set the response controls in the Customise Active listening dialog.
  5. Click Save.
Response controls Respond to custom names People often address the app by a name instead of tagging it, for example, “Hi Aria, what is the leave policy?” When you select Ignore messages that tag others, a Respond when called by custom names card appears below it. A message that contains one of these names reaches the agent, even when the message also tags someone else.
  • The card starts with the bot name configured for the app. You can remove it like any other name.
  • Add up to 10 names, including the bot name. Type a name and press Enter. Each name appears as a chip; select x on a chip to remove it.
  • The field trims spaces and rejects empty names. It also rejects duplicates, ignoring case, and shows “This name is already added.”
  • Matching ignores case and matches whole words only. “Aria” matches “Hi Aria” but not “malaria.”
  • A name can be a phrase. The agent uses the noun in the phrase as the match word. For “hey kora help me”, the match word is “Kora”: “Hi Kora” reaches the agent, but a message that contains only “help” does not.
  • Ignore small talk, Ignore off-topic questions, and Confidence check still apply to messages that contain a name.
Clearing Ignore messages that tag others hides the card but keeps the saved names. Select it again to restore them. Select the qualifier model A qualifier model scans incoming messages and decides which ones qualify for a response. Select this model once on the main Active Listening screen, and it applies to all channels. By default, the platform uses the Kore Qualifier Model.
Only one app can listen actively in a channel. If another app already has Active Listening turned on in a channel, you cannot turn it on for this app in the same channel.

Adding the App to Channels

All Slack users can add the app to any Slack channel, regardless of whether they have channel customization permissions.
  1. Open the Slack channel where you want to add the app.
  2. Click the channel name to open channel details.
  3. Click Integrations > Add an App.
  4. Search for your app and click Add.
When a user adds the app to a channel, all channel members see a confirmation message. The app follows the default behavior you configured in Publish Settings until someone with channel governance permissions customizes the channel.

Channel Behavior by Agent and Reply Mode

The app behavior in channels depends on the agent type and reply mode you configure. Search Agent with Private Reply The user who triggers the query sees an interim processing message that only they can view. After processing completes, the full response appears (without streaming), visible only to that user, along with a Post to Thread button and Search Results. When the user clicks Post to Thread, the response posts publicly in the thread. The sender can continue asking follow-up questions in the same thread, and all follow-ups remain private and contextually linked to the original interaction. Autonomous Agent with Private Reply The user who triggers the query sees a private response that only they can view. The sender can continue asking follow-up questions in the same thread. All follow-ups remain private and map to the same conversation context.
Private Reply mode doesn’t support follow-up queries from a different user in the same Slack thread. Only the original sender can continue the conversation.
Search Agent with Public Reply Autonomous Agent or Active Enterprise Knowledge Source with Public Reply

Attachments

The Slack app processes uploaded attachments as context, matching the file support of the web application. Attachment handling depends on the agent configured for the channel or the direct message. Supported Files Where Attachments Apply
  • Direct messages: The app takes attachments into context in a new conversation and within an existing thread.
  • Channels: The app processes attachments when it listens actively and when a user tags it directly. The public or private response follows the mode configured for that channel.
Files that users upload in Slack sync to the corresponding board thread in the AI for Work web and mobile applications. Error Messages Error messages appear in the public or private mode that the channel uses.
Attachment processing requires an Autonomous Agent that supports attachments. A channel or direct message configured with a Search Agent does not process attachments. Slack always requires an agent binding, so the web application behavior of uploading attachments without selecting an agent does not apply.
Active Listening Edge Case In a channel, if a user uploads a file in an unsupported format or above the size limit and does not tag the app, active listening ignores the message silently and shows no error message.

Artifacts and Thoughts

The Slack app renders generated artifacts directly in the channel, so users can view them without opening the web application. The app displays thoughts in a plan view inside the same message as the response. It does not post a separate loading message, so the thoughts and the streamed response appear together in a single message.

Rich Templates

Agent Platform agents can return rich templates, such as lists, tables, charts, forms, carousels, and media. The Slack app renders these templates in direct messages and channels. For the templates that agents can return, see Rich content.

Slack Agent Experience

The Slack app uses the Slack agent experience, which replaces the deprecated assistant view.

Migrate an Existing App

Apps that you created with an earlier manifest keep the assistant view until you migrate them.
  1. Go to the Slack Developer Portal.
  2. Select the app that you want to migrate.
  3. Go to Agents.
  4. Click Update Now.

User Access Management

Only users who have onboarded to AI for Work can interact with the Slack app. Before you deploy, complete the following:
  1. Invite all intended users to the Platform.
  2. Onboard users with the appropriate access permissions.
  3. Confirm that users exist in both the Platform and the target Slack workspace.
Users in the Slack workspace who haven’t onboarded can see the app but receive an error when they try to send a message.

Web/Mobile SDK

AI for Work integrates with your web or mobile application through the Web/Mobile SDK, enabling AI chat, agent interactions, and enterprise knowledge access directly inside your product. Each SDK app you create has its own credentials, allowed domains, and authentication settings, so you can deploy independent integrations for different applications or environments.

When to Use

Architecture

The integration uses signed credentials and JWT-based authentication between the host application and the Platform:

Prerequisites

Configuration

Each SDK app is created and managed independently from the SDK Apps page. You can create as many SDK apps as you need, each with its own credentials, allowed domains, and SDK setup. Step 1: Create an SDK App
  1. Log in to the Admin Console.
  2. Navigate to Channels → Web/Mobile SDK.
  3. On the SDK Apps page, click + Create New.
The new app opens to the configuration view with three sections in the left pane: App Credentials, Allowed Domains, and SDK Details.
You can return to the SDK Apps page at any time to view, edit, or delete existing apps.
Step 2: Configure App Credentials The App Credentials section provides the values your host application needs to authenticate with the Platform.
  • Open the App Credentials section.
  • Configure or copy the following values:
  • Click Continue.
Treat the Client Secret like a password. Regenerating the secret invalidates the previous value, and any host application using the old secret stops working until you update it.
Step 3: Configure Allowed Domains The Allowed Domains section restricts which domains can load the SDK. The system blocks requests from any domain not on the list for security.
  1. Open the Allowed Domains section.
  2. Enter a domain in the Enter domain field (for example, https://example.com).
  3. Click + Add Domain.
  4. Repeat to add additional domains.
  5. To modify the list, edit any existing domain inline, or delete a domain to remove it.
  6. Click Continue.
The system permits only configured domains to load the SDK. Add every domain where the SDK will be embedded, including staging and development environments.
Step 4: Set Up the SDK The SDK Details section provides the embed code and links to the prebuilt SDK libraries.
  1. Open the SDK Details section.
  2. Use the following resources to set up the SDK in your host application:
    • Web SDK — Click Download SDK to access the prebuilt Web SDK library and embed it in your web application.
    • Mobile SDK — Click Download SDK to access the prebuilt Mobile SDK library and embed it in your mobile application.
  3. Click Save.
Auto-Save Behavior All changes across App Credentials, Allowed Domains, and SDK Details are saved automatically. You don’t need to perform a manual save for each section — moving between sections preserves your changes.

Managing SDK Apps

From the SDK Apps page, you can:
  • View all SDK apps, including the creation date and the user who created them.
  • Edit an existing app to update credentials, allowed domains, or SDK setup.
  • Delete an app that’s no longer needed using the actions menu.
Deleting an SDK app immediately revokes its credentials. Any host application using those credentials stops functioning.

Web SDK Integration through CDN

After creating an SDK app in the Admin Console, integrate the Web SDK chatbot widget into your web application using the CDN approach. This method embeds the widget directly into your website without installing any npm packages. Integration Flow The end-to-end integration follows five stages:
  1. Load the Web SDK through CDN.
  2. Authenticate the user (through SSO).
  3. Generate a JWT token.
  4. Exchange the JWT for an access token.
  5. Initialize the chatbot widget.
Integration Prerequisites Before starting, gather the following from the SDK app you created in the Admin Console, along with details from your host application:
Never expose the Client Secret in production frontend code. In production, JWT generation must happen on the backend so that the secret stays server-side.
Step 1: Add Required CSS and JS Files Add the following inside the <head> section of your HTML page:
Step 2: Define Configuration Declare the SDK configuration values in your script. Replace YOUR_CLIENT_ID and YOUR_CLIENT_SECRET with the values from your SDK app.
Step 3: Fetch the Logged-in User Profile The SDK requires user details to complete SSO. Call your host application’s user profile API to retrieve them.
The response must include at least the user’s email, first name, and last name:
Script Placement for Steps 4–6 Place all code from Step 4 through Step 6 inside a single <script> tag in the page footer, immediately before the closing </body> tag. Don’t split these steps across multiple script blocks.
Step 4: Generate the SSO JWT Token Sign the JWT using the HS256 algorithm and the Client Secret.
In production, move JWT generation to your backend so that the Client Secret never reaches the browser.
Step 5: Exchange the JWT for an Access Token Send the signed JWT to the SSO login endpoint and retrieve the access token and user ID.
Step 6: Initialize the Chatbot Widget Initialize the widget with the access token and user ID returned from the SSO exchange. Once initialized, the chatbot automatically appears on the page.
Complete Bootstrap Example The following function combines all steps into a single bootstrap routine:

Testing Checklist

Before going live, verify the following:
  • The user profile API returns the expected data.
  • The JWT is generated correctly.
  • The SSO endpoint returns an access token.
  • The chatbot loads without console errors.
  • Conversation history is preserved across sessions.
  • The presence connection is successful.
Once all checks pass, the chatbot loads on the page, the user is auto-authenticated, chat history is preserved, and SSO authentication remains secure.