AI Employee - Knowledge Base Tables

What is Table Search in Knowledge Base?

Table Search introduces a new "Table Source" type to your CRM's Knowledge Base. By uploading CSV files, you can semantically index structured data so your AI agents can answer questions about customers, inventory, transactions, or any other tabular information during conversations.

Key Benefits of Table Search

• Natural-language queries: Ask plain-English questions about your table data without complex formulas or filters
• Semantic search intelligence: Get accurate, context-aware answers from relevant rows and columns
• Large CSV support: Handle up to 50,000 rows and 500 columns, selecting the 20 most relevant for indexing
• Structured data access: Surface tabular information that web pages and documents don't capture effectively
• Bot enablement: Empower AI employees to handle customer records, product catalogs, KPIs, and similar tabular use cases

How to Set Up Table Search

1. Navigate to AI Agents → Knowledge Base
2. Edit an existing Knowledge Base or create a new one with a name and description
3. Click Add Source and select Tables
4. Upload your CSV file using drag-and-drop or file picker (maximum 50 MB)
5. Review detected columns and adjust data types if needed
6. Click Done to begin indexing—progress will show chunking status
7. Attach the Knowledge Base to your AI Agent (Chat, Voice, or Workflow AI) as usual
8. Test with sample questions like "Which customers have overdue invoices?" to confirm results

CSV File Requirements

• Format: .csv files only (UTF-8 encoding recommended)
• Size limits: Up to 50,000 rows and 500 columns per location (select the 20 most relevant columns), with upload files up to 50 MB
• Header row: First row must contain column names
• Data types: Automatic schema detection with 80% confidence threshold
• Clean data: Remove null values, hidden formulas, merged cells, and other formatting issues

Note: CSV files can be rejected at multiple points during processing. Even if the format is accepted, data errors might still cause issues. Inspect your CSV manually in a spreadsheet program or text editor if you encounter problems.

Semantic Search Intelligence

The CRM converts each table row into vector embeddings, enabling your AI agents to understand meanings rather than just matching exact words. This allows for queries like "Show me customers who complained about billing" or "Which orders shipped overnight last week?"

The search engine compares the user's question to every row chunk and returns the most semantically similar matches—no SQL knowledge required.

Smart Table Processing

Behind the scenes, the system:

• Detects column types (text, number, date, etc.) with 80% accuracy
• Chunks rows into groups of five (maximum 2,000 characters) for efficient indexing
• Stores chunk metadata so answers can reference the correct records

Frequently Asked Questions

Can I upload Excel (.xlsx) files?

Not currently—export or save your spreadsheet as CSV before uploading.

How soon are new CSV uploads available to bots?

Typically within a few minutes—the indexing progress bar will show when processing is complete.

Does Table Search support filters or sorting in the query?

Filtering, comparison, and sorting features are coming soon. For now, ask descriptive questions or refine with follow-up queries.

Will table data appear in the Response Info sidebar?

Yes—rows that informed the answer are cited, allowing you to verify or edit them on the spot.

Can I restrict table access to specific bots?

Yes—only bots linked to the Knowledge Base containing your Table Source can query it.

How is privacy handled for sensitive CSV data?

Table Sources inherit existing Knowledge Base security; only users with access to that Knowledge Base can see or query the data.