Getting Started with Snowflake CoWork

Snowflake for Developers/Guides/Getting Started with Snowflake CoWork

Quickstart

Getting Started with Snowflake CoWork

Dash Desai

Overview

Snowflake CoWork is the personal work agent that brings deep analysis and automation to every knowledge worker. It acts across the systems users already rely on, with built-in RBAC and governance for secure and trusted answers, all inside Snowflake's perimeter. It delivers governed, curated dashboards and shareable artifacts so insights become reusable team knowledge, turning everyday work into repeatable processes and enabling teams to share and build on each other’s insights.

Unlike standalone AI tools, CoWork is grounded in your enterprise data from day one, combining AI reasoning with Snowflake's governance layer so every answer is trusted, cited, and secure.

By the end of this guide, you will have a working Cortex Agent connected to CoWork that can answer business questions across structured sales data and unstructured support tickets, generate charts, and take actions like sending summary emails.

Use Cases

CoWork moves teams from reactive Q&A to proactive, context-aware intelligence that drives decisions and execution across various business use cases:

Sales performance analysis: Sales managers can quickly get answers to complex questions like "What were my top product sales in the West region last quarter, and why did product X outperform product Y?" while analysts can understand critical trends like "Why are support tickets increasing?" by reasoning across diverse data sources.
Enhanced research & financial insights: Enrich internal data with external sources via Cortex Knowledge Extensions, allowing financial analysts to combine portfolio performance with market news, or product managers to analyze customer feedback alongside industry reports for deeper context.
Self-service data exploration: Enable all business users to independently explore data and get immediate answers to complex questions, reducing reliance on data teams and accelerating decisions across the organization.

What You'll Learn

How to create a semantic view so Cortex Analyst can answer questions about structured data
How to create a Cortex Search service over unstructured support tickets
How to build a Cortex Agent with Cortex Analyst, Cortex Search, and a custom email tool
How to ask questions, get charts, and set up an automation in Snowflake CoWork

What You'll Build

A Sales//AI agent in Snowflake CoWork that answers questions across retail sales, marketing, social media, and support data, generates charts, and emails summaries on demand or on a schedule.

What You'll Need

Git installed
Access to a Snowflake account with the ACCOUNTADMIN role in a region where you have access to the supported models. Otherwise, you will need to enable cross-region inference (see the optional statement at the end of setup.sql).

Setup

Create database, schema, tables and load data from AWS S3

Clone GitHub repository.
In Snowsight, on the left hand navigation menu, select Projects >> Workspaces, click on + Add new >> SQL File, and copy the contents of setup.sql into it.
Execute all statements in order from top to bottom. Among other things, the script creates the Snowflake CoWork object (SNOWFLAKE_INTELLIGENCE_OBJECT_DEFAULT), grants the SNOWFLAKE_INTELLIGENCE_ADMIN role usage on it, and creates the SEND_EMAIL stored procedure used later as a custom tool.

NOTE: Switch your user role in Snowsight to SNOWFLAKE_INTELLIGENCE_ADMIN.

Cortex Analyst

This tool enables the agent to query structured data in Snowflake by generating SQL. It relies on semantic views, which are mappings between business concepts (e.g., "product name," "sales") and the underlying tables and columns in your Snowflake account. This abstraction helps the LLM understand how to query your data effectively, even if your tables have complex or arbitrary naming conventions.

In Snowsight, on the left hand navigation menu, select AI & ML >> Analyst
On the top right, confirm that the role is set to SNOWFLAKE_INTELLIGENCE_ADMIN, click on Create new down arrow and select Create new Semantic View
- Location: DASH_DB_SI.RETAIL
- Name: SALES_AND_MARKETING_DATA
- Description: Semantic view for sales and marketing data analysis across campaigns, products, transactions, and social media engagement.
Click on Next, then Skip (no example SQL queries needed)
Select the DASH_DB_SI.RETAIL schema and the MARKETING_CAMPAIGN_METRICS, PRODUCTS, SALES, and SOCIAL_MEDIA tables, then click on Next
Select all available columns, then click on Next
Review and accept the suggested relationships and metrics, then click on Save and wait for the semantic view to be created

Cortex Search

This tool allows the agent to search and retrieve information from unstructured text data, such as customer support tickets, Slack conversations, or contracts. It leverages Cortex Search to index and query these text "chunks," enabling the agent to perform Retrieval Augmented Generation (RAG).

In Snowsight, on the left hand navigation menu, select AI & ML >> Search
On the top right, click on Create
- Role and Warehouse: SNOWFLAKE_INTELLIGENCE_ADMIN | DASH_WH_SI
- Database and Schema: DASH_DB_SI.RETAIL
- Name: Support_Cases
- Select data to be indexed: select SUPPORT_CASES table
- Select a search column: select TRANSCRIPT
- Select attribute column(s): select TITLE, PRODUCT
- Select columns to include in the service: Select all
- Configure your Search Service: Keep default values except select DASH_WH_SI for "Warehouse for indexing" (Choose COMPUTE_WH if DASH_WH_SI is not available)

OPTIONAL: Aggregated Support Cases using Cortex AISQL

Instead of creating a Cortex Search service for individual support cases, you may create one on aggregated support cases. This is ideal if you want the Agent to answer questions based on high-level trends rather than individual ticket details. It really depends on your use case, but it's shown here as an example using Cortex AISQL.

Execute the following SQL statements that use AI_AGG() to create aggregated support cases summary which is inserted into a new table AGGREGATED_SUPPORT_CASES_SUMMARY. Then, a Cortex Search service is created on that table.

-- Use AI_AGG to aggregate support cases summary and insert into a new table AGGREGATED_SUPPORT_CASES_SUMMARY

use database DASH_DB_SI;
use schema RETAIL;

create or replace table AGGREGATED_SUPPORT_CASES_SUMMARY as
 select 
    ai_agg(transcript,'Read and analyze all support cases to provide a long-form text summary in no less than 5000 words.') as summary
    from support_cases;

-- Create Cortex Search service on table AGGREGATED_SUPPORT_CASES_SUMMARY

create or replace cortex search service AGGREGATED_SUPPORT_CASES 
on summary 
attributes
  summary 
warehouse = dash_wh_si 
embedding_model = 'snowflake-arctic-embed-m-v1.5' 
target_lag = '1 hour' 
initialize=on_schedule 
as (
  select
    summary
  from AGGREGATED_SUPPORT_CASES_SUMMARY
);

NOTE: These SQL statements may take 3-5 minutes to run.

Create Agent

An agent is an intelligent entity within Snowflake CoWork that acts on behalf of the user. Agents are configured with specific tools and orchestration logic to answer questions and perform tasks on top of your data.

Note that you can create multiple agents for various use cases and/or business teams in your organization.

In Snowsight, on the left hand navigation menu, select AI & ML >> Agents
On the top right, click on Create agent
- Schema: SNOWFLAKE_INTELLIGENCE.AGENTS
- Agent object name: Sales_AI
- Display name: Sales//AI
Select the newly created Sales_AI agent and click on Edit on the top right corner and make the following updates.

Add Instructions

Add the following starter questions under Example questions:

Show me the trend of sales by product category between June and August
What issues are reported with jackets recently in customer support tickets?
Why did sales of Fitness Wear grow so much in July?

Add Tools

Tools are the capabilities an agent can use to accomplish a task. Think of them as the agent's skillset and note that you can add one or more of each of the following tools.

Tools
- Cortex Analyst
  - Click on + Add
    
    Add: Semantic view DASH_DB_SI.RETAIL >> SALES_AND_MARKETING_DATA
    
    Name: Sales_And_Marketing_Data
    
    Description: The Sales and Marketing Data semantic view in DASH_DB_SI.RETAIL schema provides a complete view of retail business performance by connecting marketing campaigns, product information, sales data, and social media engagement. The view enables tracking of marketing campaign effectiveness through clicks and impressions, while linking to actual sales performance across different regions. Social media engagement is monitored through influencer activities and mentions, with all data connected through product categories and IDs. The temporal alignment across tables allows for comprehensive analysis of marketing impact on sales performance and social media engagement over time.
    
    Warehouse: Custom >> DASH_WH_SI
    
    Query timeout (seconds): 60
- Cortex Search Services
  - Click on + Add
    
    Search service: DASH_DB_SI.RETAIL >> Support_Cases
    
    ID column: ID
    
    Title column: TITLE
    
    Name: Support_Cases
  NOTE: If you optionally created AGGREGATED_SUPPORT_CASES Cortex Search service, you may add it here as well.
- Custom tools
  - Click on + Add
    
    Resource type: procedure
    
    Database & Schema: DASH_DB_SI.RETAIL
    
    Custom tool identifier: DASH_DB_SI.RETAIL.SEND_EMAIL()
    
    Name: Send_Email
    
    Warehouse: DASH_WH_SI
    
    Parameter: body
    
    Description: Use HTML-Syntax for this. If the content you get is in markdown, translate it to HTML. If body is not provided, summarize the last question and use that as content for the email.
    
    Parameter: recipient_email
    
    Description: If the email is not provided, send it to the current user's email address.
    
    Parameter: subject
    
    Description: If the subject is not provided, use "Snowflake CoWork".
Orchestration Instructions: Whenever you can answer visually with a chart, always choose to generate a chart even if the user didn't specify to.
Access: SNOWFLAKE_INTELLIGENCE_ADMIN

NOTE: On the top right corner, click on Save to save the newly updated Sales_AI agent.

Snowflake CoWork

PREREQUISITE: Successful completion of steps outlined under Setup.

Open Snowflake CoWork and make sure you're signed into the right account. If you're not sure, click on your name in the bottom left >> Sign out and sign back in. Also note that your role should be set to SNOWFLAKE_INTELLIGENCE_ADMIN, the warehouse should be set to DASH_WH_SI, and your agent should be set to Sales//AI.

Now, let's ask the following questions. Answers are generated live, so wording, numbers, and chart styles will vary from run to run. The notes under each question describe what to look for.

Q1. Show me the trend of sales by product category between June and August.

What to expect: The agent uses the Sales_And_Marketing_Data tool to query the semantic view and returns a chart of sales by product category for June through August. Fitness Wear stands out with a sharp jump in July (roughly three times June's sales).

Q2. What issues are reported with jackets recently in customer support tickets?

What to expect: The agent uses the Support_Cases search tool and summarizes the matching tickets. They point to one recurring problem: ThermoJacket Pro jackets tearing along the seams, which customers describe as a manufacturing defect. The answer cites the tickets it used.

Q3. Why did sales of Fitness Wear grow so much in July?

What to expect: The agent combines sales, marketing, and social media data to explain the spike. The main driver it finds is influencer activity: Fitness Wear social mentions rose several-fold in July, led by an influencer who posted about the category only that month. Expect a chart and a short explanation of how each data source supports the conclusion.

Q4. Send a summary email

NOTE: The agent calls the Send_Email tool and confirms that the email was sent. Check your inbox for the summary email, which goes to the current user's email address unless you specify another one.

Q5. Send me this summary every Monday morning

This turns a one-time answer into an Automation that re-runs the question on fresh data and emails you the results. You can review and manage it from the Automations tab.

What to expect: The agent confirms the schedule and the first delivery date, and the new automation appears under Automations with its schedule. From there, or by asking in chat, you can edit, reschedule, pause, resume, or delete it.

Go Further

CoWork includes more capabilities you can try with the same agent:

Deep Research: Select + in the message bar >> Deep Research and ask an open-ended question like "Which product categories show the strongest link between marketing spend, social mentions, and sales, and why?" to get a multi-step, cited report.
Artifacts: Save charts and tables from any answer so you can share and revisit them without regenerating.
User skills: Codify a repeatable workflow once and invoke it with /.
Document generation: Turn analysis into PDF, Word, PowerPoint, or HTML reports. This requires adding the code execution tool to the agent.
File upload: Attach a CSV, PDF, or image directly in the chat for additional context.

Clean Up

To remove the objects created in this guide:

In Snowflake CoWork, open Automations and delete the automation you created in Q5.
Run the following SQL statements. Dropping DASH_DB_SI also removes the semantic view, Cortex Search services, and SEND_EMAIL procedure created in this guide.

USE ROLE ACCOUNTADMIN;

DROP AGENT IF EXISTS SNOWFLAKE_INTELLIGENCE.AGENTS.SALES_AI;
DROP DATABASE IF EXISTS DASH_DB_SI;
DROP WAREHOUSE IF EXISTS DASH_WH_SI;
DROP NOTIFICATION INTEGRATION IF EXISTS EMAIL_INTEGRATION;
DROP API INTEGRATION IF EXISTS SNOWFLAKE_LABS_SI_GIT_API_INTEGRATION;
DROP ROLE IF EXISTS SNOWFLAKE_INTELLIGENCE_ADMIN;

-- setup.sql changed your default role and warehouse; reset them as needed
-- ALTER USER <your_user> SET DEFAULT_ROLE = <role> DEFAULT_WAREHOUSE = <warehouse>;

NOTE: This leaves the SNOWFLAKE_INTELLIGENCE database and the SNOWFLAKE_INTELLIGENCE_OBJECT_DEFAULT CoWork object in place, since other agents in your account may use them.

Conclusion And Resources

Congratulations! You've successfully built a Cortex Agent for Snowflake CoWork that can respond to questions by reasoning over both structured and unstructured data.

What You Learned

You've learned how to create the fundamental building blocks for Snowflake CoWork. This Cortex Agent is now a powerful asset, capable of combining deep analysis and automated tasks right from a single chat interface.

Related Resources

Updated Oct 20, 2026

This content is provided as is, and is not maintained on an ongoing basis. It may be out of date with current Snowflake instances

Ready to get started?

Open in Snowflake