All 47 cards in Doing read with their comments and due dates, then handed back with a proposed destination for each and the evidence beside it.
Forty-seven cards sit in Doing. Nine have not moved in three weeks, three are the same request typed by three people, and Done has 200 cards nobody has archived since spring.
A companion reads all 47 with their comments and due dates and answers the question a filter cannot: is this card stuck, finished, or was it never a card. You approve the moves you agree with.
01
What does “stale” actually mean on your board?
A card with a due date two weeks past and a last comment saying “shipped, closing this” belongs in Done. A card with no description and no comments belongs in a conversation. A filter cannot tell those apart; reading the comments can.
What comes back is a short list with a proposed destination for each card and the evidence beside it. You tick the moves you agree with, so nothing shifts under the other four people watching.
02
Cards written from the thing that caused them
The description is thin because whoever filed the card was in a hurry.
A companion reads the source, the customer message, the meeting notes, the page you are on, then writes what is being asked and what done looks like.
Then it attaches the link back to the source, so a card opened in three weeks still explains itself. Feeding a board from research done elsewhere is the same shape as data extraction, with cards as the destination.
03
How do you get the right name on the card?
The board’s members are listed first, so “give this to Sam” resolves to the Sam on this board rather than a name in a prompt. The same goes for labels: they are read off the board, so a companion can tell you which cards are mislabelled.
Assignment and comments both count as things the team will see, so both are staged for you.
04
Decide which cards are actually finished
Which of these 47 cards is finished.
Which two are duplicates. What this card should say so it still makes sense in three weeks. Each needs somebody to read the comments and decide, and that is the reading a companion takes off you.
Butler, as of this writing, runs the deterministic rules on Trello’s servers: archive a Done card after 14 days, create the standup card every Monday. The judgement pass above is the one that stops the board accumulating.
Yes. Trello is a native integration. A companion lists boards, reads lists and cards, creates cards, updates name, description, due date and list, archives or deletes, comments, attaches a URL, and assigns board members. Changes to a shared board are staged for your approval.
It reads the whole list and proposes the rest. Cards come back sorted into finished, genuinely stuck, and never-really-a-card, with one batch of moves for you to approve.
The card tools cover name, description, due date, list, archive state, comments, URL attachments and member assignment, and a board’s labels are readable, so a companion tells you which cards are mislabelled. Applying the label is a click you keep.
Point a companion at the message or the notes. It writes a card whose description explains the request, the current state and what done looks like, attaches a link back to the source, and assigns a member from the board’s real member list. You see the card before it is created.
A companion covers the judgement work: triage, deduplication, and writing cards from the email or notes that caused them. Download Strawberry free, then see getting started for describing a pass once and reusing it.