
Turn Your CSV Data into Accurate Answers
You can now explicitly guide your AI on how to handle your data tables, ensuring that every search is pinpoint accurate and every answer is perfectly formatted.
This feature is available within the Functions configuration for all CSV-powered bots.
What’s New
- Targeted Query Generation (Input Columns): Manually designate which columns the AI should use for filtering (e.g., searching by
hospital_nameorcity). - Precision Output Mapping (Output Columns): Define exactly which data points should be extracted and returned to the user (e.g., prioritize
bed_countandaccreditation). - Enhanced AI Context (Column Hints): Add "semantic hints" to headers so the LLM understands exactly what the data represents (e.g., "This column stores the hospital's specialty").
- Dynamic Schema Synchronization: Automatically detect and prompt for new mapping updates whenever you replace a CSV file in your function.
What This Solves
Previously:
- The AI had to "guess" which columns were relevant for a search
- Broad, inefficient queries sometimes led to data mismatch or hallucinations
- Admins had no way to prioritize specific data points in the final response
This led to:
- Occasional incorrect search results for complex tables
- AI providing "too much" irrelevant data in the answer
- Difficulty handling tables with ambiguous or technical column names
Now, admins provide the "map," and the AI follows it with 100% precision.
How to Use
- Open your AI agent and navigate to the Functions tab in the sidebar.
- Click to edit an existing CSV Function (or create a new one).
- In the modal, locate the Input Columns (optional) section.
- Select the columns you want the AI to search by from the dropdowns.
- Add a Description (hint) for each column to help the AI understand its context.
- Click Update/Save to apply the intelligent logic instantly.
---
Why This Matters for Your Business
Zero-Hallucination Search
By limiting the AI's search scope to the correct columns, you ensure the answers are always grounded in your actual data.
Cleaner, Professional Answers
Don't let the AI dump the whole row. Control exactly which "details" your customers see for a more polished experience.
Better Handling of Complex Data
Hints allow the AI to master technical tables (like ERP or Inventory exports) that would normally confuse a standard chatbot.
Key Benefits
- 100% Search Accuracy — No more "guessing" which column to filter
- Semantic Understanding — Hints bridge the gap between technical headers and human questions
- Reduced Token Usage — More efficient queries mean faster and cheaper responses
- Total Control — You decide exactly what data the bot is allowed to show
For Users
- Faster and more accurate answers to data-specific questions
- Cleaner, more relevant responses without extra "data noise."
- Higher confidence in the bot's ability to handle complex lookups
Summary
Intelligent CSV-to-SQL Configuration eliminates guesswork by allowing admins to explicitly map data roles and provide semantic hints within the Functions tab, ensuring 100% accuracy and cleaner responses for data-heavy interactions.
Ready to try these new features?
Experience the latest improvements and see how they can enhance your workflow. Get started today or learn more about what's coming next.
More Changelog Updates
Related Blog Posts

Build WhatsApp Lead Qualification Bot | Smart Automation | Kipps.AI
Build a WhatsApp lead qualification AI agent with Kipps.AI. Automatically filter, score, and route quality leads in real time—no human intervention needed.
Read more
Voice AI for Lead Qualification | Voice Agent | Kipps.AI
Use Kipps.AI to build voice-based AI agents for lead qualification. Automate calls, ask smart questions, and route hot leads instantly.
Read more
Lead Qualification AI Agent for Zoho CRM | Kipps.AI
Create a powerful AI agent that qualifies leads and syncs them into Zoho CRM automatically with Kipps.AI.
Read more