Map with AI reads the first few rows and works out which column is which — useful when the headers are a spreadsheet somebody else made (“Assigned To”, “Deadline”, “Project”). Columns this app already recognizes by name are matched without it and are marked matched by name; the AI only fills the gaps and can never overrule them. Every field is a dropdown, so change anything that looks wrong before importing. It needs the API key from the task app's Settings, and nothing is imported until you press Import.

First row must be a header. Works with a hand-made file or a raw tasks table export from Supabase — column names are matched case-insensitively and both spellings are accepted (due date or due_date, external link or jira_link, and so on).

Imported: title, description, owners, assignee, contributors, status, due_date, completed + completed_date, topic / topic_id, external link / jira_link, is_weekly_agenda, is_okr, archived, tag / tags. Only a title or description is actually required.

Swimlanes, topics and tags are created as needed. A topic column names the topic group a task belongs in, a tag column names its tags (one per cell, or several separated by ; — commas are kept as part of the name, so “Build, buy, or partner” stays one tag), and a status that isn't one of this board's swimlanes gets a new lane appended. Anything this board doesn't already have is added before the tasks are imported, and the preview lists them by name first. Matching is case-insensitive, so a lane, topic or tag that already exists is reused rather than duplicated. New swimlanes change the board's columns for every member, which is why they're called out first in the preview.

Not carried over: id and user_id — a newly imported task gets a new id and is owned by you (the database only permits writing tasks you own), and an updated one keeps the owner it already has. board_id is ignored; tasks always land on the board you're viewing. starred, the email_* columns and source_note_id are personal-app fields this board doesn't store.

Matched by description — updated, never duplicated. A row whose description already exists on this board updates that task instead of adding a second copy; rows with a blank description match on title instead, and a description repeated inside the file itself is only used once. An update only writes the columns your file actually has, and a blank cell leaves the existing value alone — so a partial file (say description + status) is a safe way to bulk-edit. Tags are added to whatever the task already carries rather than replacing them. Tasks owned by other members can't be updated from here and are reported as skipped.

A row with a blank status goes to the first lane (there is no name to create a lane from), and a row marked complete never creates a Complete lane — it goes to the Completed tab instead; a topic_id from another board is ignored (the topic label is used instead), and a row with neither stays in “No topic”. Rows marked complete or archived go straight to those tabs rather than onto the swimlanes.