Prospect tables, rows and AI columns

Where rows come from, what an AI column is, how to fill cells, and how to mark a table ready.

A prospect table is the list of sites your agents pitch. This is where finding turns into rows you can review, enrich, and approve.

Tables belong to the workspace#

A table belongs to the workspace, not to one agent, so "every agent in the workspace works every table". Create one with New table: "Name it, and say how we should find the rows that go in it." You also choose a required How we find rows method: Article Search, Links From Pages, or "work it out from the brief".

Where rows come from#

Rows arrive four ways:

  • Find leads / Find articles. The agent runs a source plan and pulls in a batch, with a count you set from 1 to 200.
  • Upload CSV. Bring your own list.
  • Add Row. Add a single site by hand.
  • From chat. Ask the agent or the manager to find and add rows.

Each row is a candidate site or page, and it carries its own details, draft, and status chip.

Row status at a glance#

A row shows one of these chips: Email Sent, Follow Up Sent, Sending, Failed, No email, Duplicate, Suppressed, Customer, Replied, or Bounced (a hard bounce outranks every stage below it). You also set some yourself: Not ready, Ready to send, and Booked. The table tabs group rows by stage: All, Drafts, Ready to Send, Email Sent, Follow Up Sent, and Replied, each with a count.

AI columns#

An AI column fills its cells with a prompt you write, so you can enrich every row at once. Add one with the + in the header row. The editor asks for:

  • Column name and a Prompt. Type @ to insert another column's value, for example: "Return a clean, official version of this company name (no LLC/Inc): {company_name}".
  • Format and Type, with Options (one per line) if the type is a select.
  • Compute now, which runs the prompt on every row as soon as the column is created. Leave it off and the column starts empty, to be filled later.

Filling and refilling cells#

  • Compute now runs the prompt across the table immediately.
  • Enrich fills only the empty and failed cells: use it to top up without touching good data.
  • Re-run overwrites cells, including ones that already have values: use it when you changed the prompt.

What to do when a cell is empty#

An empty cell usually means the agent could not find or compute a value for that row. You can:

  • Enrich the empty and failed cells to try again.
  • Fix the source data (for example a bad URL) and re-run.
  • Leave rows with No email out of your approvals, since they cannot be sent anyway.

Marking a table ready#

Approval happens per row. Mark a row Ready to send to authorize it, or use the bulk Mark ready to send and Un-mark actions. Nothing is emailed until a row is ready, and the backend refuses a row that is not.

Note: if you see "No agent works this table yet, so there is nobody to mark these ready for", assign an agent to the workspace first. A table with no agent has nobody to send its rows.

The table kebab menu holds the rest: Rename table, Edit brief ("What belongs in this table"), Clear data, Export as CSV, Copy table ID, Duplicate table, and Delete table. Once rows are drafted and ready, move to approve and send.