Natural-language query
Ask for the data in plain English. Exponam Analyst Intelligence translates the question into SQL, shows you the query before it runs, executes it on your Databricks or Snowflake SQL engine, and returns the result to your worksheet. The SQL is yours to read, edit, and rerun.
Ask, read, run
Example: “Show me all intercompany transactions between the US parent and the Irish subsidiary for 2022–2024, by month, by cost center.” The generated query appears in the SQL scratchpad. Run it as written, or adjust a filter first.
Natural language
Question in, SQL shown, results in the worksheet.
What you need
Excel for Windows
With the add-in installed from the public install page. Name and email are requested; no payment.
Platform access
A Databricks or Snowflake account with access to a SQL warehouse, using the credentials you already hold.
A model for the translation step
One of two, described below. Model selection and configuration are covered in the Getting started guide.
The translation step
A local model
Running on your own machine. No question, schema, or token leaves your network. Larger models translate better and run slower; Exponam publishes sizing guidance from 8B to 32B parameters.
Your organization’s commercial model
Under your own API agreement. Exponam sends the model three things: your question, a fixed set of translation instructions, and schema metadata – table names, column names, and data types. It sends no data. No rows, no values, no credentials, and nothing to Exponam. The exposure is the same as sending the model a table definition.
If your organization already permits use of that model, this path introduces no new category of information; whether table and column names may leave the network is a question your data-classification policy already answers. A commercial model also returns a plain-language description of the result alongside the SQL.
Either way, no row of your data leaves your perimeter.
Cost and governance
What it costs the platform
The query runs on the platform’s SQL engine (Databricks or Snowflake) and is metered the way any query from that warehouse would be. Nothing routes through Exponam.
Governance
Access is the platform’s. If your credentials cannot see a table in Databricks or Snowflake, the add-in cannot return it to Excel. Every result is tied to a visible query, so a number in a workbook traces to the SQL that produced it and can be reproduced on demand.
For the perimeter question, our Security Center documents data flow and the NIST SP 800-218 attestation.
Trying it
Install, connect a workspace, choose a model, and ask. Results land at up to one million rows per query without a license; enterprise licensing extends the ceiling to ten million rows and adds MSI deployment. The built-in sample Delta Share exercises the visual data builder, not this path. Natural-language query needs a connected workspace.
The other two paths: Direct SQL and Delta Sharing import.