Automate repetitive work with agents
Sigma Assistant answers questions and builds on governed data. Sigma Agents act on what changes, in Sigma or from Claude and ChatGPT.
One version of the numbers
Sigma Assistant, agents, and Claude read the same data models and metrics, so answers match the dashboard.
Answers that finish the job
Agents write back to input tables and call your APIs, with a person’s sign-off when you want one.
Safe to give to everyone
Answers follow the permissions you already manage in Sigma and your data warehouse.
Trusted by 2,000+ leading enterprises around the world
One governed answer
Your team is already using AI on your data
Most of it runs on exports, in chat windows your data team can’t see. Sigma puts that AI on the data you already govern.
| Today | With Sigma |
|---|---|
| A CSV pasted into a chat window for each summary | Assistant, agents, and Claude on one set of data models |
| Two people ask the same question and get two numbers | Answers that match the dashboard because they query it |
| Anyone with the file can see each row in it | Queries that run as the person asking |
| Recurring summaries written by hand | Agents that write the summaries and take the next step |
Everything an agent needs to finish the job
Ask and build with Sigma Assistant. Hand the work that repeats to Sigma Agents, in Sigma or from the AI assistants your team already uses, on one governed foundation.
Ask and build
Ask a question or describe what to build
Ask Sigma Assistant a question from the home page or any workbook, or describe the dashboard you want and build it together.
- Steps and sources behind each answer
- A plan to review before anything gets built
- Answers from the data models your team certified
Build agents
Build an agent in plain language
Describe the job in plain language and get a draft agent to review, with its instructions and data sources already filled in.
- Instructions, data sources, and actions in one place
- Call warehouse agents, search services, and MCP connectors
- Promote from dev to prod with version tags
Act in chat
Answers that change the record with sign-off
Add an agent to a workbook or AI App as a chat element. Review each change it proposes and approve it in chat before anything is written to your input tables or sent to your APIs.
- Writeback to input tables in your data warehouse
- REST API calls, webhooks, and stored procedures
- Approve or skip each change in chat
Run headless
Run agents on a schedule, from Claude, or from code
Put an agent on a schedule, or call it from Claude or your own code. Each run stays inside the caller’s permissions.
- Call agents from Claude through the Sigma MCP server
- REST API with streaming and structured output
- Runs logged with source, tokens, and latency
Connected to your AI
Bring Claude, ChatGPT, and Codex
Connect Claude, ChatGPT, and Codex through the Sigma MCP server and plugin. Find, query, and build workbooks and data models from those assistants, through each person’s Sigma login.
- Each person’s permissions, through OAuth
- The same numbers as the dashboard, not a pasted export
- Build and edit workbooks from Codex and ChatGPT
Reply…
Start from an agent someone already built
Open a working example from the library and point it at your own data.
IT Ticket Triage AgentA Sigma agent reads each ticket, calls a Snowflake Cortex warehouse agent, and writes the triage decision back to the ticket row.
Claims Anomaly Detection AgentA claims analyst pulls live data, drills with a spreadsheet UI, writes back changes, and chats with an agent that summarizes the case.
Expense forecast agentAn expense forecast agent collects parameters, analyzes historical trends, and generates growth assumptions, while the workbook surfaces approval-ready memos.
Capacity Planning Sales AgentModel rep ramp, quotas, and territory changes on live data. A scenario-planning Sigma agent does the math, the workbook routes the approvals.
Agents that follow the rules you already set
Each agent follows the permissions you set in Sigma, and your data warehouse enforces its own policies underneath.
Human approval
Require sign-off in chat before an agent changes any data.
Your choice of model
Admins pick the AI provider for the org: models in your data warehouse, or OpenAI, Gemini, Amazon Bedrock, and more.
Cost you can see
Admins see questions, active users, and tokens per agent, and can sort to find the costliest.
Per-user OAuth
Connect with OAuth so queries run with each person’s own warehouse role.
Row- and column-level security
Reads and writes both inherit your table policies.
Audit trail
Audit logs record user activity, AI conversations, and input table edits, so you can see who changed what.
Teams putting AI to work on Sigma
See how customers bring AI to governed data with Sigma
500+ business users empowered with real-time insights. 15-16 terabytes of compressed data in dashboards. $20 billion assets under management.AB CarVal Harnesses AI in Sigma to Unlock Secure, Real-Time Investment Insights
Weeks to days to ship new workflows. 14K+ r&d orgs on luma. 2M+ scientists using the data.How Dotmatics ships new workflows in days with Sigma
“Using Sigma, I was able to make a dashboard in under two hours that got the answers. The AI features let our team move faster than ever before.”How Duolingo Democratized Data Access While Protecting Sensitive InformationStraight from the source
“Sigma allows me to set up an AI agent to answer questions about my company’s incentive plan.”
JamesEnterprise · G2 (opens in a new tab)“The recently added Sigma Assistant feature lets end users query the data in the reports and dashboards in plain language, and that makes it super easy and convenient for non-technical users.”
EngineerSoftware · Gartner Peer Insights (opens in a new tab)“Sigma has been quick to adapt powerful AI solutions within the modern data stack, from consuming AI-powered data sets to using LLMs to leverage AI at the BI layer directly.”
Senior Data AnalystSoftware · Gartner Peer Insights (opens in a new tab)“Sigma’s AI assistant has become a complete game-changer for ad-hoc requests.”
Cherry LinPrincipal Data Analytics, GTM · Enterprise · G2 (opens in a new tab)“They even have a formula assistant where you can type out what you want to do and their AI will build out the formula for you to use.”
Data AnalystTransportation · Gartner Peer Insights (opens in a new tab)“From a product perspective, they try to be on the cutting edge, offering numerous AI features and workbooks as code.”
Kyle WeberLead BI Developer · Enterprise · G2 (opens in a new tab)“We like that it’s not just a BI platform, it’s more of a workflow platform.”
Andrew DennisDirector of Data · Mid-Market · G2 (opens in a new tab)“Reports are now live, actions can be taken immediately, and collaborating with colleagues is faster and more accurate than it has ever been.”
Alex BurcumSr. Data Analyst · Small Business · G2 (opens in a new tab)
Before you build your first agent
Sigma Agents are now GA: Scale AI workflows without the sprawl
Run Sigma Agents from Claude, ChatGPT, your own apps, or a Sigma workbook, while admins track each agent's usage, access, and token spend.
Read the blog: Sigma Agents are now GA: Scale AI workflows without the sprawlSigma Agents: See our Agents in Action
Join for a live look at what’s new and what you can now do with Sigma Agents as a builder.
Watch the webinar: Sigma Agents: See our Agents in ActionAI Agents for Sales & RevOps: What They Do, Key Use Cases and How to Implement Them
Learn how AI agents for sales and RevOps work, key use cases like pipeline risk and approval routing, and how to implement them with governed warehouse data.
Read the blog: AI Agents for Sales & RevOps: What They Do, Key Use Cases and How to Implement Them
Sigma agents FAQ
The questions worth asking before putting AI agents on your data.
What is a Sigma agent?
A Sigma agent is an AI agent your team builds in plain language on live data in your data warehouse. It can answer questions or run on a schedule, and it can write back or call other systems within the user’s permissions.
Can we start with human approval and move to autonomy later?
Yes. Most teams start in chat, approving each change, then move an agent onto a schedule once its output holds up.
Which AI model runs our agents?
Your admins choose the AI provider as an org setting: models hosted in Snowflake, Databricks, or BigQuery, or external providers like OpenAI, Azure OpenAI, Gemini, and Amazon Bedrock. Agent queries run on your data warehouse compute either way.
What does an agent cost to run?
Agents use consumption credits. Admins see questions, active users, and token usage per agent over the last 7 to 90 days, and can sort to find the highest-cost agents.
Do Sigma agents work with Claude or ChatGPT?
Yes, and many teams use both. The Sigma MCP server lets Claude and ChatGPT find the workbooks and data models your team already built and query them through each person’s Sigma login.
How are Sigma Agents different from agentic analytics?
Agentic analytics, like Sigma Assistant, helps people ask questions and build. A Sigma agent also captures what’s happening in your data and acts on it, like writing back or opening a ticket, with a person’s sign-off if you want one. Agents run in Sigma or from Claude, ChatGPT, and your own code.
When should we use Sigma Assistant, an agent, or Claude?
Use Sigma Assistant to ask questions and build inside Sigma. Use an agent when a team needs the same job done the same way every time. Use Claude or ChatGPT through the Sigma MCP server when people want to stay in the AI tool they already use. All three read the same governed data.
What is a headless agent?
An agent that runs without a chat window. An automated action calls it on a schedule, or your code calls it through the REST API, and it sends its results by email or Slack or writes them back to an input table.
Stop being the team everyone waits on
Bring the request that keeps landing on your desk, and we’ll show you how the people asking could run it themselves on governed data.