An AI data analyst lets a marketer ask a question such as "what was our new-customer revenue from Meta last month?" in plain language and get an answer from the company's own data. The tool writes the SQL, runs it against the warehouse and returns a table or chart. For an ecommerce brand, the three main options are Snowflake Cortex Analyst, Databricks Genie and Gemini in BigQuery, and the right one is almost always the one built into the warehouse the brand already uses. This guide compares what each does and costs at published prices, and sets out what has to be in place before the answers can be trusted. It is from Brand Media Works, an AI consulting and marketing data engineering firm for ecommerce brands.
How do Cortex Analyst, Databricks Genie and Gemini in BigQuery compare?
| Snowflake Cortex Analyst | Databricks Genie | Gemini in BigQuery | |
|---|---|---|---|
| Works on | Data in Snowflake | Data governed in Databricks Unity Catalog | Data in BigQuery |
| How the AI is billed | Through Cortex Agents in token-based AI credits, or the standalone API per 1,000 messages | In DBUs. Each user gets 150 DBUs a month free, and 150 DBUs equal $10.50 in US East | SQL and Python code assist, data canvas and data preparation at no additional cost |
| Query compute | Running the generated SQL is billed as normal warehouse compute | Genie compute, such as SQL Serverless, is billed separately | Queries are billed as normal BigQuery usage |
| How analysts improve answers | A semantic view of tables, metrics, synonyms and verified example questions | Datasets, sample queries, instructions, metrics, business rules and verified answers | Data agents with custom metadata, instructions, glossary terms and verified queries |
Databricks also says Genie One and Genie Agents usage by users is free until 31 January 2027, and applies a 25% promotional discount to billed usage. Google bills BigQuery queries on demand at $6.25 per TiB after the first 1 TiB each month, which is free (read 18 October 2026).
What does an AI data analyst cost to set up, and when is a custom build worth it?
The software is the cheap part. Each vendor charges little or nothing to try the AI layer, and the queries it writes cost the same as any other query. The cost is in the data underneath:
- Getting the data in. Shopify orders, ad spend from every platform and email results have to be loaded into the warehouse and refreshed. One data consultancy publishes about $4,000 for custom pipelines from one or two sources with automated refresh (read 25 October 2026).
- Agreeing the definitions. Revenue, new customer, return and margin each need one definition the tool can use. All three vendors rely on analysts adding definitions and verified example queries for this.
- Testing the answers. A set of known questions with known answers, checked every time the definitions change.
A custom-built AI analyst is worth it in a narrower set of cases: when data sits in several warehouses or tools that none of the vendor products reach, when the team wants answers inside Slack or another tool with its own permissions, or when the analyst needs to take actions as well as answer questions. Published build prices for that kind of agent are higher: one AI firm puts a data analysis agent at $80,000 to $180,000 to build (read 18 October 2026). For most brands, the better path is the vendor tool on top of a well-modeled warehouse. Custom AI tools describes how Brand Media Works scopes this work, and how to build a marketing data warehouse covers the foundation.
How can I use Gemini in BigQuery to analyze data?
- Write SQL from a prompt. In the BigQuery editor, describe the query in a comment or use the SQL generation tool, and Gemini writes the SQL or Python.
- Explore with data canvas. Data canvas lets a user find, join and query tables, visualize results and share the work with others, using natural language.
- Start from data insights. BigQuery suggests queries generated from a table's metadata on its Insights tab. Google says data insights use an org-level quota and are available on on-demand compute and the Enterprise editions.
- Build a data agent. For repeat questions, create a data agent with the brand's metadata, instructions, glossary terms and verified queries, and let the team chat with it.
The core features need no extra license. Google's separate Gemini Code Assist licenses, at $19 to $54 per user per month depending on edition and commitment, are not required for them (read 2 November 2026).
What should a marketing team check before trusting the answers?
- One revenue number. Does the tool's revenue match the store's net sales for the same week?
- Attribution rules. Is "revenue from Meta" the platform's own count, a model, or a test result? The tool will not know unless it is told.
- The SQL. Someone who reads SQL should review the queries behind new kinds of question before they become a habit. Snowflake notes that Cortex Analyst handles only questions SQL can answer and can struggle in long or shifting conversations.
- Access. Restrict customer-level data with warehouse permissions, not in the prompt. Snowflake says Cortex Analyst follows its role-based access controls and does not train on customer data, and Databricks says Genie answers are governed through Unity Catalog.
Frequently asked questions
What is an AI data analyst?
A tool that answers questions about a company's data in plain language by writing and running queries against its warehouse, then explaining or charting the result. It is also called text-to-SQL or conversational analytics.
Which AI data analyst tool is best?
The one that runs where the data already is: Cortex Analyst on Snowflake, Genie on Databricks and Gemini on BigQuery. Moving data to a different warehouse for its AI tool rarely pays off.
Is Gemini in BigQuery free?
Its core features, SQL and Python code assist, data canvas and data preparation, are available in all BigQuery editions at no additional cost. The queries it runs are billed as normal BigQuery usage.