Skip to main content

Operations Knowledge Base

The Operations Knowledge Base is a shared, tenant-wide repository for internal documentation that all Specialist Agents can access.

Overview​

Unlike per-agent knowledge stores, the Operations Knowledge Base is a centralized resource for all SMEs:

  • Shared access: Any SME can query relevant information
  • Organized by domain: Documents tagged by department and category
  • AI-powered search: Uses Azure AI Search for intelligent retrieval
  • Access control: SMEs see documents appropriate to their role

Architecture​

+------------------------------------------+
|       Operations Knowledge Base           |
+------------------------------------------+
|                                          |
|  HR Documents    Accounting    Engineering|
|  - Handbook      - Expenses    - API Docs |
|  - Benefits      - Invoicing   - Standards|
|  - PTO Policy    - Tax Proc    - Runbooks |
|                                          |
|  Marketing       Sales         Legal      |
|  - Brand Guide   - Pricing     - Contracts|
|  - Templates     - Outreach    - Policies |
|                                          |
+------------------------------------------+
                    |
            All SMEs can query
           based on access level

Document Categories​

Documents are organized into categories for efficient retrieval:

CategoryDescriptionExamples
PolicyOfficial company policiesExpense Policy, PTO Policy
ProcedureStep-by-step processesOnboarding Checklist, Invoice Process
ReferenceReference materialsAPI Documentation, Code Standards
TrainingTraining materialsNew Hire Guide, Tool Tutorials
TemplateDocument templatesReport Templates, Email Templates
GuideHow-to guidesBrand Guidelines, Style Guide

Domain Tags​

Documents are also tagged by department:

  • hr - Human Resources
  • accounting - Finance/Accounting
  • engineering - Technical/Engineering
  • marketing - Marketing/Brand
  • sales - Sales/Business Development
  • legal - Legal/Compliance
  • operations - General Operations

Uploading Documents​

Supported File Types​

TypeExtensionsMax Size
PDF.pdf10 MB
Word.docx, .doc10 MB
Text.txt, .md5 MB
PowerPoint.pptx10 MB
Excel.xlsx5 MB

Upload Process​

  1. Navigate to Build > Knowledge > Operations Knowledge
  2. Click Upload Document
  3. Select or drag the file
  4. Configure metadata:
    • Title: Display name for the document
    • Category: Select from policy, procedure, reference, etc.
    • Domain Tags: Tag with relevant departments
    • Associated SME: (Optional) Primary SME for this document
  5. Click Upload

Processing Pipeline​

After upload, documents go through automatic processing:

  1. Text Extraction: Content extracted from the file
  2. Chunking: Document split into searchable segments
  3. Indexing: Chunks indexed in Azure AI Search
  4. Embedding: Vector embeddings generated for semantic search

Processing typically takes 1-5 minutes depending on document size.

SME Access Levels​

Control which SMEs can access which documents:

Own (Default)​

  • SME only sees documents specifically uploaded for it
  • Most restrictive setting
  • Use for sensitive, domain-specific documents

Department​

  • SME sees documents in its domain and related domains
  • Allows cross-functional information access
  • Use for SMEs that need broader context

All​

  • SME sees all tenant documents
  • Use for executive-level or administrative SMEs
  • Requires careful consideration of data sensitivity

Querying the Knowledge Base​

SMEs use the query_operations_knowledge tool to search:

SME: "What is our expense approval process?"

Tool call: query_operations_knowledge
{
    "query": "expense approval process",
    "filter_by_category": "procedure"
}

Result:
- Document: "Expense Reimbursement Policy"
- Section: "Approval Thresholds"
- Content: "Expenses under $100 require no approval..."

Query Parameters​

ParameterDescriptionExample
querySearch question or keywords"vacation policy"
filter_by_domainLimit to specific domain"hr"
filter_by_categoryLimit to document category"policy"
filter_by_tagsMatch specific tags["benefits", "employees"]
max_resultsNumber of results (1-10)5

Managing Documents​

Viewing Documents​

  1. Navigate to Build > Knowledge > Operations Knowledge
  2. Browse documents by category or search
  3. Click a document to view details and indexed content

Document Details​

The detail view shows:

  • Metadata: Title, category, tags, upload date
  • Processing Status: Indexed, processing, or failed
  • Chunk Count: Number of searchable segments
  • Access Statistics: Query count, last accessed

Updating Documents​

To update an existing document:

  1. Upload a new version with the same title
  2. The old version is replaced after processing
  3. Index updates automatically

Deleting Documents​

  1. Find the document in the list
  2. Click Delete
  3. Confirm deletion
Deletion is Permanent

Deleted documents are removed from the index and cannot be recovered. Consider archiving instead of deleting if you might need the document later.

Analytics and Monitoring​

Dashboard Metrics​

The Operations Knowledge dashboard shows:

  • Total Documents: Count of indexed documents
  • Total Chunks: Number of searchable segments
  • Storage Used: Total size of uploaded documents
  • Queries (30 days): Recent query volume

Query Analytics​

Track how the knowledge base is being used:

  • Most frequently searched terms
  • Documents with highest hit rates
  • SMEs with most queries
  • Failed queries (no results)

Recommendations​

Based on analytics, the system may recommend:

  • Documents that should be updated (frequently searched but low satisfaction)
  • Gaps in coverage (queries with no results)
  • Redundant documents (similar content)

Provisioning​

The Operations Knowledge Base is automatically provisioned when:

  1. Your tenant has Azure OpenAI configured
  2. Your tenant has Azure AI Search configured
  3. SME Agents feature is enabled

Manual Provisioning​

If not auto-provisioned:

  1. Navigate to Build > Knowledge > Operations Knowledge
  2. Click Provision Knowledge Base
  3. Select your Azure providers
  4. Wait for index creation (1-2 minutes)

Best Practices​

Document Organization​

  1. Use clear titles: Descriptive names help SMEs find documents
  2. Tag consistently: Use the same tags across related documents
  3. Keep documents current: Outdated information causes problems
  4. Single source of truth: Avoid duplicate documents with conflicting information

Content Quality​

  1. Write for search: Include keywords users might search for
  2. Use headings: Structure documents for easy chunking
  3. Be specific: Vague content doesn't answer specific questions
  4. Include examples: Concrete examples improve understanding

Access Control​

  1. Review access levels: Ensure SMEs have appropriate access
  2. Sensitive documents: Use "Own" access for confidential information
  3. Audit regularly: Check who has access to what

Troubleshooting​

Document Not Found in Searches​

  1. Check processing status (may still be indexing)
  2. Verify document category and tags match the query filters
  3. Ensure the querying SME has appropriate access level
  4. Try different search terms

Processing Failures​

If a document fails to process:

  1. Check file format is supported
  2. Verify file isn't corrupted
  3. Ensure file size is within limits
  4. Try re-uploading the document

Slow Queries​

If queries are slow:

  1. Check Azure AI Search service status
  2. Review query complexity (too many filters)
  3. Consider reducing max_results