Skip to main content
Enterprise Knowledge is a workspace-level fallback search layer. When no agent handles a user query, the system automatically searches configured knowledge sources and returns a contextual response. Supported source types:
  • Search AI – indexes and retrieves from document repositories and structured data
  • Agentic App – connects to action-oriented services for retrieval and workflow tasks
  • Amazon Q – integrates with Amazon Q Business applications for enterprise search
Enterprise Knowledge is available in enterprise workspaces only.

​
How It Works

  1. A user submits a query.
  2. The system checks for a matching agent (action, agentic flow, or retrieval agent).
  3. If no agent matches, Enterprise Knowledge activates automatically.
  4. The active knowledge source searches configured content and retrieves relevant results.
  5. The system returns a response with citations and matching search results.

​
Configuration Limits

SettingLimit
Maximum sources5
Active sources1 at a time
Source typesOne type per configuration (Search AI, Agentic App, or Amazon Q)
Configuration is per workspace — each enterprise workspace maintains its own knowledge sources. Configured sources are available to all users in the selected workspace. For restricted access, select the workspace that matches the required user access level.

​
Configure a Knowledge Source

Go to Enterprise Knowledge > Configure > Create New, then select the source type.

​
Connect Search AI

Search AI indexes content from document repositories and structured data sources for precise retrieval. Before you start: Gather the following credentials from your Search AI app. In the Search AI app, go to Manage > Dev Tools > Web/Mobile SDK. See Search AI Credentials for details.
FieldDescription
Source NameUnique name for this knowledge source
DescriptionPurpose and contents of this source
URLEndpoint URL for the Search AI service
App IDApplication ID of the Search AI app
Client IDClient credentials for RAG Agent interaction
Client Secret IDSecret key for secure interaction
Search IDUnique client identifier from Search AI
Required API Scopes Associate these four API scopes in your AI for Work platform app channel:
  • Answer Generation
  • Permission Entity Management
  • Document Management
  • Facets
After connecting:
  1. Go to Enterprise Knowledge > AI Search Agents > Go to Search AI to access the associated app.
  2. Configure content sources in Search AI.
  3. Fine-tune configurations for your responses.
  4. Enable Answers.
  5. Train the application.
The Admin Console shows an overview of connected content sources including uploaded files, indexed webpages, and third-party connectors with sync status.

​
Connect Agentic App

Agentic App connects to services that can both retrieve information and perform actions on behalf of users. Steps:
  1. Go to Enterprise Knowledge > Configure > Create New > Agentic App.
  2. Define the API connection:
FieldDescription
URLEndpoint URL from the Agentic Platform
MethodHTTP request type (GET, POST, etc.)
HeadersAuthentication tokens and request metadata; add multiple as needed
Query StringURL parameters to filter or customize requests
Body TypeData format (JSON, Raw, etc.)
Content TypeData type label for the API
Request ContentData payload for the API request
CURL Import (optional): To auto-populate fields from an existing API, click Curl Import, paste your CURL command, and click Import.
  1. Click Save.

​
Connect Amazon Q

Amazon Q integrates with Amazon Q Business applications to provide enterprise search across AWS-managed content. Prerequisites:
  • Account Admin, Workspace Admin, or Workspace Collaborator role
  • Active Amazon Q Business applications and retrievers in AWS
Steps:
  1. Go to Enterprise Knowledge > Configure > Create New > Amazon Q.
  2. Enter a Source Name and Description.
  3. Copy the Tenant ID displayed — required when setting up the data accessor in AWS.
  4. Enter the following details from AWS:
FieldDescription
Application IDUnique identifier of your Amazon Q Business application
Retriever IDUnique identifier of your Amazon Q Business retriever
Access Resource NameARN for secure access to Amazon Q resources
Application LocationAWS region where your Amazon Q Business application is deployed
IDC LocationAWS region where your AWS Identity Center instance is located
  1. Click Save.
  2. Set the source as Active to enable it for queries.
When Amazon Q is the active source, all non-agent queries route to the Q Index. Enterprise Workspace is the default access scope for all users; you can select a different workspace during setup.

​
Work – Agentic Search

Work – Agentic Search is an intelligent search and orchestration layer across your organization’s knowledge sources. It provides source attribution, comprehensive citations, and document export (PDF, PowerPoint).

​
Prerequisites

  • Admin credentials for the Agent Platform
  • OpenAI or Azure OpenAI API credentials
  • Platform admin panel access with appropriate permissions
  • Active session tokens for MCP tool authentication

​
Step 1: Create the Agentic Application

  1. Download the base app export package from the link provided by your administrator.
  2. In the Agent Platform, go to Create App > Import Existing App.
  3. Upload the base app file.
The import establishes your working environment with core components pre-configured. You’ll customize these in the following steps.

​
Step 2: Configure Language Models

  1. Go to Models > Add Model and add your OpenAI or Azure OpenAI models.
  2. Open Supervisor and select a model.
  3. Open Data Assistant Agent and select a model.
You can use the same model for both agents or different models based on performance requirements.

​
Step 3: Set Up MCP Tools

MCP tools connect Agentic Search to your enterprise systems using session-based authentication. Get the Server URL:
  1. In the Platform, go to Admin Hub > Assistant Configuration > MCP Server > Default Server.
  2. Copy the Server URL.
Create the MCP Tool:
  1. In the Agent Platform, go to Tools > MCP Tools > Create Tool.
  2. Paste the Server URL.
  3. Add the authentication header:
    • Key: auth
    • Value: {{memory.sessionMeta.metadata.aiForWork.sessionToken}}
  4. Save the tool.
Enable Tool Integrations:
  1. Go to Tools > Tool Selection.
  2. Enable the tools your organization uses:
    • Google Connector (Contact Lookup, Send Email, Read Emails, Calendar)
    • Outlook tools
    • Active Search Apps (based on your tenant configuration)

​
Step 4: Configure Agent Workflow

How you configure the workflow depends on how many search apps your organization uses. Multiple search apps: Keep the Query Hub agent. It routes queries to the appropriate search app. Add all search apps to Query Hub and configure their descriptions in Knowledge Source settings. Single search app: Delete the Query Hub to simplify the workflow. Queries flow directly to your search app. Configure Data Assistant:
  1. Open the Data Assistant agent.
  2. Add all enabled tools to this agent to give it access to all necessary integrations.

​
Step 5: Create and Deploy a Version

  1. Go to Deployment > Versions > + New Version and save.
  2. Go to Deployment > Environments > + New Environment.
  3. Deploy your version to the new environment.

​
Step 6: Generate API Credentials

  1. Copy the deployment cURL from your environment.
  2. Go to API Scoping > + New API Scope and fill in the required fields.
  3. Go to API Keys > Create Key using your scope.
  4. Replace the placeholder API key in the cURL with your generated key.
Keep the updated cURL command ready for the next step.

​
Step 7: Register with AI for Work

  1. In the Platform, go to Knowledge Sources > Agentic Apps.
  2. Enter a name and description for your application.
  3. Paste the updated cURL command.
  4. In the JSON body, update the streaming configuration:
    • From: "stream": {"enable": false}
    • To: "stream": {"enable": true}
  5. Click Save.
Enabling streaming activates real-time token-level response display for a better user experience.

​
Validate the Deployment

  1. Go to AI for Work > Work tab > Search Bar.
  2. Submit a test query.
  3. Verify results appear correctly.
  4. Check Agent Platform > Traces to review the execution flow.
Traces show how the query moved through the agent architecture, which tools were invoked, and how the response was generated. Use this view to troubleshoot issues.

​
Agent Configuration Guidelines

Use these guidelines to ensure reliable, consistent Agentic Search behavior.
GuidelineDetails
Define agent scopeRestrict each agent to Agentic Search responses only. Prohibit use of general AI knowledge not originating from configured search apps.
Enforce source citationsRequire citations in every response for traceability and auditability.
Handle out-of-scope queriesConfigure agents to explicitly acknowledge when information is unavailable rather than inferring answers.
Write detailed app descriptionsIn Enterprise Knowledge Source settings, include each app’s purpose, data sources covered, and scope boundaries.
Test scope complianceAfter configuration, test with out-of-scope and edge-case queries to verify agents respect defined boundaries.
Structure promptsUse XML-style tags (e.g., <section>...</section>) and Markdown (headings, bold, blockquotes) to organize instructions by priority.
Handle edge casesIn the Supervisor prompt, define logic for response formatting, data volume limits, and output aggregation.
Dynamic Context Placeholders Use these placeholders in agent prompts to include real-time user and session context: 
{{memory.sessionMeta.metadata.firstName}}                                          // User first name
{{memory.sessionMeta.metadata.lastName}}                                           // User last name
{{memory.sessionMeta.metadata.email}}                                              // User email
{{memory.sessionMeta.metadata.timezone}}                                           // User timezone
{{memory.sessionMeta.metadata.customData.entrKnwAgnAppMeta.userProfile}}           // User profile
{{memory.sessionMeta.metadata.customData.entrKnwAgnAppMeta.currentDateAndTime}}    // Current date and time

​
Manage Knowledge Sources

Access source management at Enterprise Knowledge > Configure.
ActionHow to
ActivateClick ⋮ next to an inactive source > Set Active. The previously active source deactivates automatically.
EditClick ⋮ > Edit to update parameters, content references, or response settings without creating a new source.
DeleteClick ⋮ > Delete. This action is permanent and cannot be undone.