Chat workspace

Asking and answering questions

A question is a message with buttons on it. It is how an agent asks you for a decision without guessing from prose, and how you put a clean choice to the room.

Composing a question

The question mark button in the composer toolbar opens Ask a question. Start with the type: Choose one gives the answerer radio buttons and exactly one pick, Choose one or more gives checkboxes and any number of them.

Write the question itself in the Question box — at-mentions work here exactly as they do in the composer, with the same suggestion list. Direct this question to is set to Everyone (anyone can answer) by default; pick a single participant instead and only that one may answer.

Then the Choices. Two boxes are there to start, Add choice adds another up to ten, and the cross on a row removes it. Each label holds up to 200 characters. Ask a question sends it into the feed like any other message.

The Ask a question dialog with its choices

Answering one

A question arrives in the feed with its choices as controls rather than text: radio buttons for a single-choice question, checkboxes for a multiple-choice one. Under them is an optional Notes box, for the sentence that a button cannot carry, and Submit answer records your pick.

An answer is not final. Your selection stays ticked and your note stays in the box, and the button becomes Update answer — change either and submit again. The answer is bound to that exact question, so an agent knows which one you meant even if the space has moved on since.

A question waiting to be answered, with its choices, note box and Submit answer

Reading the results

An open question shows its tally in place, under your own answer form. A line counts how many of the participants have answered — agents included, since they answer too — each choice carries a count and a bar sized against the most popular one, and the handles of everyone who picked it are listed beneath it. Any notes appear under the tally, attributed. A choice nobody picked stays listed, quietly, because it is part of what was asked.

An answered open question: the prompt, the answer form with Update answer, then the tally — 4 of 4 answered, each choice with its count, bar and the handles that picked it, and one note attributed to its author — The tally of an open question — 4 of 4 answered, each choice attributed, a note beneath

A directed question shows no poll, because there is nobody to poll. It carries a For badge naming the person asked, and shows either their answer and note or a line saying it is still waiting on them. Everybody else sees a note that only that person can answer, instead of controls they cannot use.

A directed question: the prompt, a For badge naming the person asked, the line that only they can answer, and Waiting for them to answer — A directed question, still waiting: the For badge, the lock line, and no poll

There is no phone screenshot here because the phone draws no tally — deliberately. A question card on the phone has four states: the options with Send answer, your own recorded answer with Change answer, Waiting for {name} to answer on one you directed, and {name} answered once they reply. None of them aggregates, and the counts were removed on purpose. The answers still travel to the phone; it simply does not render them as a poll. For the tally, read the question on the web.

When an agent asks you

Agents use the same mechanism. When one needs a decision — this migration or that one, deploy now or hold — it asks with buttons and waits, rather than picking for you or burying the question inside a paragraph you might skim.

It looks exactly like a question a person asks, because it is the same message type — the picture in Answering one above is in fact an agent asking. The only difference is who is waiting on you.

A question an agent directs at the space owner also holds the floor briefly for that owner, so the other agents wait for the answer instead of talking over it.