# Run over a dataset

> Point your agent at rows instead of chats: run it across a dataset in the Enterprise workspace.

Source: https://cotera.co/docs/deploy/run-over-a-dataset

---

Chat runs one task at a time. A dataset runs your agent at scale: one run per row, results landing in columns next to your data. Classify ten thousand tickets, enrich every account, score each lead.

Dataset work lives in the **Enterprise** workspace, the classic data side of Cotera. Same account, same org, same agents; a different surface, built for tables.

## Switch to the Enterprise workspace

1. Press `Cmd+K` (or `Ctrl+K`) anywhere in Coco and type "switch".
2. Choose **Switch App Version**.
3. Pick **Enterprise** and click **Switch Version**. The app reloads into the classic workspace.
4. Switch back the same way at any time. Nothing is lost in either direction.

![](https://cotera.co/media/docs/deploy/deploy-switch-dialog.png)

If you don't see the option, the classic workspace isn't enabled for your org yet. Ask us and we'll turn it on.

## Work with datasets

The Enterprise workspace organizes around **Datasets**, **Agents**, and **Documents**:

![](https://cotera.co/media/docs/deploy/deploy-enterprise-home.png)

1. [Create a dataset](https://cotera.co/docs/guides/datasets/getting-started.md): from a CSV upload, a warehouse connection ([Snowflake](https://cotera.co/docs/reference/data-warehouse/snowflake.md), [BigQuery](https://cotera.co/docs/reference/data-warehouse/big-query.md)), or [EraQL](https://cotera.co/docs/guides/how-to-create-a-dataset-using-eraql.md).
2. Add an intelligent column: an [LLM column](https://cotera.co/docs/guides/how-to-create-an-llm-column.md) runs a prompt per row; [agents in datasets](https://cotera.co/docs/guides/datasets/agents-in-datasets.md) run a full agent (tools and all) per row.
3. Run it. Each row becomes one run, and results fill the column. Start with a handful of rows before running thousands.

## When to use which surface

1. **A question, a task, a build:** chat with Coco.
2. **The same job across many records:** a dataset.
3. **Your data already lives in Snowflake:** [run from Snowflake](https://cotera.co/docs/deploy/run-from-snowflake.md) and skip the import entirely.