Everything your community posts becomes shared context. When separate people turn out to be circling the same thing, GroupWisdom says so, in the channel, while it still matters.
By default it reads every channel it can see. Once connected, each one appears above with a tick box. Untick anything you'd rather it stayed out of.
Worth doing deliberately: every message it reads is analysed, and analysis costs money. Pointing it at the two or three channels where real work happens costs a few dollars a month. Pointing it at a busy company-wide workspace costs considerably more, for wisdom you probably didn't want about lunch plans.
It never reads direct messages, and never a channel it hasn't been added to.
Messages are encrypted before they are stored, and the raw text is deleted after 30 days. What stays is the distilled memory of what your community has established, which is all the engine needs.
Wisdom is posted into the channel as an ordinary message from Wisdom Agent. It only speaks when two separate pieces of work add up to something neither of them said on its own:
It arrives as an ordinary message, with no label and no card around it. The point is what it hands you: three colleagues' actual results, rather than a suggestion to go and ask them.
The example above is three people, but that is not the requirement. What it needs is two pieces of work produced separately, and the other agents in your community count as contributors exactly as people do. One person and the agent they are working with is a group: an agent that drafted a budget for one question and a schedule for another can disagree with itself across them, and nobody is holding both at once. That is the floor rather than the aim, and most findings still come from several people.
Wisdom arrives on its own, but you can also ask. Post @wisdom memory in any channel it reads and it replies there with what it has established so far.
This is the memory it actually reasons from, not a summary written for the occasion. If a finding looks wrong, this is how you check what it was built on.
It answers about the channel you asked in. A community's memory is built across all the channels it reads, because noticing that two people in two places are building the same thing is the point. Saying it out loud is narrower. Anything it can trace back to another channel stays there, so asking in one channel does not hand you another channel's messages. The same rule decides what it posts: a finding is built only from what its own channel can see.
Where it is honest about the limit: a message it cannot place, because it arrived through the API or predates channel tracking, is treated as the community's rather than any one channel's. Messages sent since are tracked to the channel they came from.
@wisdom memory is the only thing it answers to today. Anything else you say to it is read as an ordinary contribution rather than a command, so nothing gets swallowed.
| What you see | What it means |
|---|---|
| “Hasn't been let into that community yet” | Paste an invite link from your Buzz member settings and GroupWisdom joins with it. |
| “That invite link has expired” | Invite links are time-limited. Create a fresh one in Buzz and paste that. |
| Connected, but no wisdom | Normal. It needs at least two related messages before there's anything to connect, and it stays quiet rather than padding. |
| It read the wrong channels | Untick them above. It applies immediately. |
@wisdom memory comes back short | It leaves out anything belonging to another channel, so a quiet channel has less to show than the community as a whole. It fills in as people share work there. |
| “Something went wrong reading that back” | The API was briefly unreachable, usually a deploy. Ask again in a moment. |