Brain questions & giving feedback

The Brain is built to be honest about what it doesn't know. Rather than guess, it sometimes asks you a question - and your answers are one of the most powerful ways to improve it.

Why it asks questions

When the Brain spots something it can't resolve from the data alone - an ambiguous result, a gap in context, a hypothesis it can't confirm - it raises a question instead of inventing an answer. This keeps its reasoning trustworthy: surfacing uncertainty rather than hiding it.

Answering

Questions appear on the dashboard (and in Today). For each one you can:

• Answer it - your response is folded into the Brain's knowledge with high confidence.
• Dismiss it - if it's not useful or not relevant, clear it away.

Answering tends to have an outsized effect, because you're filling exactly the gap the Brain identified.

Questions don't wait forever. The Brain keeps a short list - about five at a time - so the card never becomes a backlog. A question nobody has answered after 30 days is cleared on its own, and once one has been waiting two weeks a newer question can take its place. Nothing is lost when that happens: if it still matters, the Brain asks again. Reviews, and points you raised yourself that are waiting for a yes, are never cleared this way.

A review of earlier records is settled right there. When you have told the Brain that something replaces what it knew, the item shows the earlier records with tick boxes: confirm to retire the ticked ones, or use Undo this review afterwards to put them back. See Telling the Brain something is out of date.

The same kind of review appears when you turn down a document point the Brain had already used, or replace a document with a newer version. Records that simply copied the point are retired with it; anything the Brain worked out from it is listed here instead, because it may rest on other things as well. Records that rest only on what you turned down arrive ticked, the rest unticked, and nothing is retired until you confirm.

Not every item is a question to answer. Some are reviews: a claim someone flagged as wrong, or a set of older records to retire after you told the Brain something replaces them. Those are settled by choosing what to do rather than by typing a reply, so they show no answer box - open the claim they are about and decide there. Typing an answer on one would only add a new note beside the thing you were trying to settle, so the Brain declines it and says where to go instead.

When the question came from something you said in the chat, you don't have to come back here for it. The point appears under the Brain's reply marked "Flagged to confirm", with a Save it anyway link that confirms it on the spot. See Chatting with the Brain.

Giving feedback elsewhere

You also teach the Brain just by using the platform - setting goals, recording lost-deal reasons, adding comments on cards, and uploading documents. It reflects on all of it.

A comment on a number is remembered against that number, not just the card's title: which channel and metric it was, the date range you were looking at, and the figure on screen at the time. So a note like "we paused that campaign on the 12th" on the sessions tile is there the next time sessions for that month come up, whether you ask in chat, in a report, or through a connected AI assistant.