Demo: Cortex Analyst on Snowflake
Platform: Snowflake
Feature: Cortex Analyst (Natural Language to SQL)
Presenter: Awaish
Focus: Creating semantic views and querying data using natural language to eliminate analyst bottlenecks
Last Updated: 2026-02-16
Demo Script
WHAT: Introduction
Hi, my name is Awaish and today we’re demoing the Cortex Analyst feature on Snowflake’s platform.
WHY: Value Proposition
This feature is important because it solves the problem of data bottlenecks in analytics workflows. Traditionally, when business users or analysts need data insights, they have to wait for data engineers or SQL experts to write queries, which can take hours or even days. This creates delays in decision-making and slows down business operations.
With Snowflake’s Cortex Analyst, you can create domain-specific AI analysts that understand your data and answer questions in plain English. Each analyst becomes an expert in a specific business area—like wholesale, retail, or e-commerce—and can instantly generate SQL queries from natural language questions. This eliminates wait times, empowers teams to be self-sufficient, and accelerates the time from question to insight.
HOW: Feature Walkthrough
Here’s how it works:
1. Create a Semantic View (Your Domain-Specific Analyst)
First, I’ll show you how to set up an analyst. In Snowflake, you can see I have Cortex Analyst here. We can add as many semantic views as we want—each semantic view is like a separate analyst that’s expert in a specific domain.
I’ve already built a view for wholesale, but let me show you how to create a new one. You select Create New Semantic View, and then you select the tables and columns that will be part of that semantic view. This should be specific to some business domain—like retail, e-commerce, or you can even go more granular like Amazon-specific or Shopify-specific.
Snowflake recommends going as granular as you can because then the analyst can better answer your questions.
2. Select Your Tables and Columns
For example, let me create a retail analyst. I’ll select the schema—let’s say the retail schema. Here we have 10 tables, but we can select only the ones that are really needed to answer our specific questions.
Maybe I just want answers about sales performance, so I’ll select: sales data, product performance summary, weekly sales summary. I’m not interested in inventory or warehousing right now, so I’ll skip those.
You can select all columns from these tables, or go even deeper and select individual columns. I’ll say Create and Save.
3. Configure Relationships and Metrics
Now Snowflake creates the semantic view and automatically adds descriptions and suggests relationships between the tables. You can see all the tables listed—which ones are dimensions, which are time dimensions, which are facts.
At the bottom, there are 3 important things you can configure:
- Verified Metrics: Add business metrics like “net revenue equals total revenue minus discounts”
- Relationships: Define how to join tables together (like how to join products with sales)
- Verified Queries: Add queries you’ve already tested that work well
This helps the AI understand your business logic and data structure.
4. Ask Questions in Natural Language
Now our analyst is ready! Let me go to the Wholesale Analyst I created earlier. Here I can ask any questions in natural language.
For example: “How many partners are there in Austin, Texas?”
And here it gives us an answer—it shows the total count, and importantly, it shows the SQL query it used to answer this. We can verify if it’s correct, which tables it used, what fields it selected.
5. Iterate and Refine
I can continue the conversation. Let me ask: “What is the total revenue from these partners?”
Ah, interesting—it says it can get the total revenue from the wholesale geography table, but it can’t filter based on month and year because that data isn’t available in this table.
So I can continue prompting: “Maybe join this with the date table?” The AI will understand and refine the query.
That’s how we can continue to prompt and ask the AI to figure out which tables can be used and how to get the answer we need.
REVIEW: Recap
Thanks for watching me demo how you can create domain-specific AI analysts and query data using natural language on Snowflake’s Cortex Analyst.
This feature transforms analytics from a waiting game into an interactive, real-time exploration. Instead of waiting for SQL experts, business users can ask questions directly and get immediate answers—while still being able to verify and understand the SQL behind the scenes.
BRAND: Call to Action
At Brainforge we specialize in implementing software like Snowflake and modern data platforms. If you’re interested in learning how we can help your business set up semantic views, build AI analysts for your teams, and unlock faster analytics, please get in touch.
Key Features Demonstrated
- 🎯 Semantic Views: Create domain-specific AI analysts (wholesale, retail, e-commerce)
- 📊 Granular Control: Select specific tables and columns for each analyst
- 🔗 Relationships & Metrics: Configure business logic and verified queries
- 💬 Natural Language Queries: Ask questions in plain English
- ✅ SQL Transparency: See and verify the generated SQL
- 🔄 Iterative Refinement: Continue the conversation to refine queries
Use Cases
- Business analysts needing ad-hoc reports without SQL knowledge
- Sales teams exploring performance data by region or product
- Finance teams pulling real-time revenue and partner metrics
- Operations teams monitoring KPIs without engineering support
Key Benefits
- ⚡ Instant insights - No more waiting in queue for data requests
- 🎯 Domain expertise - Each analyst specializes in specific business areas
- 📖 Transparency - See the SQL query and verify the logic
- ✅ Accurate queries - AI understands your schema, relationships, and business metrics
- 🔄 Conversational - Refine and iterate until you get the right answer