It reads only the channels you invite it to, each as its own project, and speaks when two pieces of finished work in a channel add up to something neither said alone. Most of the time it says nothing.
The workspace is connected to your GroupWisdom account, and its findings live there.
It reads nothing until you do. In the channel, send:
/invite @GroupWisdom
It says hello, explains what it keeps, and starts reading from that moment. Each channel you invite it to becomes its own project with its own memory, so what it learns in one channel never shapes what it says in another.
A finding arrives as a reply to the message that completed it, and is also shown in the channel. Under each one are three buttons, Helpful, Wrong and Late. One press tells us whether it was worth saying, which is the only honest measure of whether this works.
/groupwisdom memory/groupwisdom mutemute today lifts at midnight UTC. It keeps reading./groupwisdom unmute/groupwisdom demoThe same words work as a mention, @GroupWisdom memory, for anyone who prefers that.
Message text, the name of the person who wrote it, and the channel, from the public channels it has been invited to. It does not read private channels, direct messages, files, or channels shared with people outside your workspace, and it never asks Slack for a channel's history: it starts from the moment it is invited.
Findings are written by an AI model and can be wrong, incomplete or late. Each finding records the words it rests on,
and /groupwisdom memory shows what it believes, so a finding can always be checked rather than trusted.
Nothing read here trains any model. The full detail is in the privacy policy.