Data Analyst Agent · Use case
How can business users query Snowflake in plain English with AI on our own infrastructure?
A data analyst agent turns a question asked in plain English into a query, runs it against Snowflake or another database, and returns the answer with the query it ran. It never runs a query that changes data. The agent and the model run on your own infrastructure, so questions and results are not sent to a public AI service.
Every question waits for an analyst.
And the data cannot leave.
- ✗ 01Business users wait for an analyst to write each query, and analysts spend their days on routine requests.
- ✗ 02A hosted assistant would need the schema, the question and the results sent outside. A figure with no query behind it cannot be checked.
Where it stops for a person
Never runs a query that changes data.The agent runs read queries only. A request to change or delete data is refused and left to a person with the right access.
It writes the query.
And shows it.
Step · 01
The agent is given a read-only database account and a description of the tables it may use.
Step · 02
A user asks a question in plain English.
Step · 03
The agent writes a query and shows it.
Step · 04
It runs the query and returns the result, as a figure or a table.
Step · 05
The question, the query and the result are logged.
Reads from · connectors
The agent connects to Snowflake through the Snowflake connector and to other SQL databases through the Database connector.
Acts through · MCP servers
- ODOracle Database
MongoDB
ClickHouse
Amazon RedshiftMetabase
- VIVizro
Through MCP servers, the agent can also query Oracle Database, MongoDB, ClickHouse and Amazon Redshift, and connect to Metabase or Vizro for charts.
What you set up
- Create a read-only database account for the agent.
- Describe the tables and columns it may use, in business terms.
- Write test questions with known answers.
Where it runs
The agent, its knowledge store and the model run on infrastructure you own or control. You choose the model and serve it yourself, so questions, documents and answers are not sent to a public AI service.
How you get it today
This page is labelled Use case: the agent is built in Katonic Studio. You describe the job, connect what is named above, add test questions and release the agent once it passes them. When the Catalog holds a tested template for this job, this page will carry the Template label and show how to take it.
Can the agent change or delete data?
No. It has a read-only account and runs read queries only.
How do users know a figure is right?
The agent shows the query behind every answer, so an analyst can check it. Each version is scored against test questions before release.
Does our data go to a hosted model?
No. The model is one you serve on your own infrastructure. Query results go from the database to the agent and no further.
Is it limited to Snowflake?
No. Other SQL databases connect through the Database connector, and MCP servers reach stores such as Oracle Database and ClickHouse.