A whole team, one AI: shared chats that feel like sitting at the same desk

Workplace comedy has a whole genre thanks to single-player AI chats. A useful answer gets screenshotted, dropped in the team channel, and the sender spends the afternoon fielding "what did you ask before that?" The context dies in transit, every single time.

Shared chats in Raven are our fix. People join by email invite or @username, and once they accept, the conversation belongs to the group. Everyone types, every message lands on every screen as it is sent, and Raven answers with the whole room in one shared context. What that is worth: the follow-up you would have asked badly gets asked well by someone else, and the spec lands while the first answer is still being read. Raven ends up playing the colleague who has somehow read everything ever written.

Names and photos ride on every bubble. Replies stream onto every participant's screen as they form, so watching a colleague's question get answered live on your laptop while they type it from their phone is usually the moment the feature clicks for people.

One thread, everyone in it

There are no forks and no "which version are you looking at", because the group shares one history: every message, every reply, in order, synced live across every device. Regenerate an answer and everyone sees the new version with the old ones a click away, which turns arguments about the answer into comparisons of the answer.

When the humans need the floor

Discussion mode hands the room to the humans for a while. Messages keep landing for everyone, but Raven sits the round out and waits, watching. The next message brings the model back, carrying everything it watched. Teams run "should we even ask this" huddles with it, or let two departments argue about requirements while the AI politely waits. And @mentions do the obvious thing: type @, pick a person, pull them into the point.

The frozen-link alternative

ChatGPT and Claude "share" a chat by sending a link to a frozen snapshot of a dead conversation. Fine for showing someone an answer. Useless for producing one together. The live room, the shared context, and the model present while the thinking happens: that is the feature, and work gets done in the meeting, not in the transcript.

Open any chat's Share menu in the Raven app, invite the people you trust with the question, and watch the answer get better because more than one brain was in the room. The permissions and presence side has its own write-up in the companion piece.