Skip to main content

Enterprise Analytics Assistant

Business Need​

Provide employees with a unified AI assistant that can both search across enterprise documents (emails, files, tickets) and query structured data (databases, warehouses) through a single conversational interface.

Solution Overview​

This solution combines enterprise search capabilities with text-to-SQL analytics in a single Custom Chatbot:

Step 1: Configure User-Level Search

  • Document Sources: User connectors provide access to Outlook, SharePoint, OneDrive, and Jira
  • Real-Time Search: No data ingestion; searches execute with user credentials respecting access controls
  • Semantic Retrieval: Finds relevant documents, emails, and tickets based on natural language queries

Step 2: Configure Text-to-SQL Analytics

  • Data Warehouse: Direct connection to Snowflake for structured data queries
  • Schema Understanding: Automatically maps business questions to database tables and columns
  • Query Generation: Translates natural language to optimized SQL queries
  • Result Visualization: Generates charts and tables from query results
💡Pro Tip

Combining enterprise search with text-to-SQL creates a powerful "unified data assistant" that eliminates the false boundary between structured and unstructured data. Users get answers regardless of where the information lives.

Step 3: Build Hybrid Custom Chatbot

A Custom Chatbot is configured with:

  • Multiple Knowledge Sources: Both user-level document connectors and database connections
  • Intelligent Routing: Automatically determines whether to search documents or query databases based on the question
  • Cross-Source Synthesis: Can combine insights from both structured data and documents in a single response
  • Context Awareness: Maintains conversation history and can answer follow-up questions

Step 4: Deploy Enterprise-Wide

The assistant is deployed through the web interface accessible via the Abacus ChatLLM Teams UI.

How It's Used in Practice​

Employees interact with the analytics assistant for a wide range of information needs:

Document Search Queries:

  • "Find the product roadmap presentation from last month"
  • "Show me all emails about the customer onboarding project"
  • "What are the open P1 bugs in Jira?"

Analytical Queries:

  • "Show me revenue by product line for Q4"
  • "Compare our sales pipeline to this time last year"
  • "What's the average customer lifetime value?"

✨Key Outcomes

Key achievements:

🔍 Single point of access​

Eliminated need to search across 5+ systems

⚡ Faster insights​

From question to answer in seconds

📊 Combined document & data analysis​

Unified unstructured and structured information

🎯 80% question resolution rate​

Minimal escalation to analysts or IT support

Additional Information​