Grok Bot 1.0 changes what happens when a colleague corrects the team's shared AI. In the beta, whatever a teammate taught the Bot was saved for everyone straight away. Now the Bot only suggests the change, and the suggestion waits for the owner or a Manager to approve or reject it.
Why the waiting list is the real news
Other tools already put an AI in Slack, and AI notetakers already joined calls. What is new here is that a team's Bot can now be taught by many people and has a gate on what it keeps. That matters because a shared Bot's memory works like a document that every answer is built from. When one person teaches it a better way to write a supplier email or check an invoice, the whole team gets it, which is the point. The same route also carries a wrong lesson to everyone: a rushed fix, a one-off exception taught as a rule, a habit that only one colleague has.
Each waiting item says what it would do, for example update a skill or forget a team memory, and the owner or Manager chooses Approve or Reject. What the Bot learns about one person stays between that person and the Bot. Each person's chat is private, even from the owner.
Wikipedia shows what a gate like this costs and what it buys. Its reviewers spend time, and in return the pages can stay open to anyone. The Bot's version is cheap. If thirty people teach it ten things a week and each item takes two minutes to read, that is about twenty minutes a week for the reviewer. It is only cheap if the reviewer knows the work.
The reviewer should be the person who would correct a new hire
The owner is whoever created the Bot, and that is often the person who was curious or the one in IT, not the person who knows how month-end close or a returns policy really works. A Bot that is shared with a whole team will drift toward what its reviewer lets through. So the person who would correct a new colleague on that job is the right reviewer, and the list becomes the place where the team's unwritten habits get written down and argued over.
Sharing a login now means the Bot acts as you
The same release lets the owner decide, app by app, whose account the Bot uses. Teammates can connect their own, the owner can share theirs, or the Bot can use a separate team account where the app allows it. Sharing the owner's account only works after a team admin turns on a switch that is off by default. Whatever the Bot does through that connector, for any teammate, happens as the owner.
That is the old habit of sharing one login with the whole office, with a Bot now doing the typing. In a Slack channel the Bot works only with the owner's apps, while in a private chat it asks before using yours. For anything with customer records or money in it, a separate team account is the safer choice.
Meetings are the smaller part
A Bot can now join Google Meet, Zoom and Microsoft Teams calls, but notetakers such as Otter and Fireflies have done that for some time. The difference is that this is the Bot the team already shares. It joins under a visible name, answers in the meeting chat instead of speaking, and posts a summary and action items afterward, so what the team taught it can come into the call.
Team Bots need a Team or Enterprise plan. Meetings also work on the Ultra plan, and someone working alone gets none of the team features.
A team that tries this should start with one Bot for one shared job and share it only with the people who do that job, as the September launch of Team Bots already allowed. It should name the reviewer before the first lesson is taught, and read the list after the first week.