Enrich a Google Sheet

Point the product at a spreadsheet, describe what you want found for each row, and it fills in the columns — on a schedule, row by row, keeping track of which rows it has already done.

1. Connect your Google account

Workspace settings → Integrations → Connect Google account.

Your connection is yours. Each member connects their own, and the card shows only your connection — a teammate’s will not work for you. Tokens are stored encrypted and refresh themselves, so a long job cannot outlive its access.

Disconnecting revokes access without deleting anything. Jobs over sheets you attached pause with a “sheet access revoked” marker, and pick up again once you reconnect and re-enable them.

2. Attach the spreadsheet

Attaching tells us which sheet, which tab, and what the columns mean:

• Key column — what identifies a row. This is how we know a row we have already handled from a new one, so pick something stable.
• Read columns — what we may look at.
• Write columns — where results go. Usually the same sheet.

Attaching uses your connection and makes you the source’s owner: only you (or the workspace owner, as a backstop) can change its mapping or detach it. Any member of the workspace can then preview and build jobs over it — that is the point of attaching. Previewing a sheet nobody has attached needs your own connection.

Detaching is a soft withdraw by default: dependent jobs pause and history survives. Deleting everything is a separate, explicit choice.

3. Describe the job

Create the job from the canvas — select the steps, then “Run per row…” — or from the jobs list.

• Instructions — plain English. “For each company, find their customer support email and HQ city from their website.” This is the goal for every row.
• How to gather — browse the web, search only, or reason over the row’s own data with no lookup.
• Cadence — how often to check for work. Fifteen minutes is the floor; presets cover the common cases. You can also run one now.

You do not define an output schema. The fields come from the write columns you mapped, so adding a mapped column is how you ask for another field — it gets filled on the next run. You can add a description per field to steer what goes in it.

Worth knowing about: a batch size, a re-check window for rows already done, a confidence threshold that parks doubtful rows instead of writing them, and the option to write results into a different sheet than the one you read.

4. Who can do what

• A job created from a workflow selection inherits the workflow’s members. Share the workflow, not the job.
• A standalone job is yours plus whoever you share it with: member can run, enable and edit; viewer can only look. People you share with are notified.
• Whoever runs a shared job, the rows run on the sheet owner’s access. Permission to run a job is not permission to reach the data behind it.

5. Following along

Each row carries a status, and the row list is the record of what happened:

Status	What it means	What to do
Pending	Queued for the next run	Nothing
In progress	Being worked on now	Nothing — a stalled row is picked up again automatically
Enriched	Written back; will be revisited after your re-check window	Nothing
Needs review	Below your confidence threshold, so not written	Review it yourself
Failed	Errored — the reason is on the row	Fix the cause; it retries on the next run

“Failed” often means your Google connection needs reconnecting.

On the canvas the job appears as a frame around its steps, with the source, cadence and live counts on it. Two jobs over the same sheet are two separate frames that each fill their own columns and keep their own row histories.

If nothing runs

• The workspace’s subscription has lapsed. Jobs are skipped, visibly, until there is an active plan.
• A new row never gets picked up. Check the key column actually has a value in it; a row with no identity cannot be tracked.
• Results are not appearing in the sheet. Check the write columns are still mapped and still exist in the tab.