Collaboration

What you can do

Threads, mentions, reactions, attachments, live co-editing and a history of what happened - with your AI employees writing in the same threads as everybody else

Comment on what you are looking at

Threads attach to the objects people argue about:

  • A record - a request, a deal, a task in a process. The question about that request sits on it.
  • A dataset - the source itself, so "this figure looks wrong" is attached to where the figure came from.
  • A range of cells in a spreadsheet - a comment on the number, not on the file.
  • A project - the thread about how it is going, where the project is.
  • A checklist item - small enough that most tools would not bother, and exactly where people argue about whether something is done.
  • A block on a page while it is being designed - pinned where the problem is, not described in a message. See Design.

Mentions that reach people

Mention somebody and it leaves the screen:

  • In their channel - Slack, Telegram or email, whichever they use.
  • In their inbox - a list of what is waiting for them, with a badge, so nothing depends on them happening to open the right page.
  • Marked as read when they read it, per mention, so the inbox is a queue and not a growing pile.

Agents write in the same threads

Your AI employees are participants here, not a separate channel of notifications.

An agent that checked a supplier, gathered the numbers, drafted a specification or noticed something wrong writes it as a comment on the record it worked on - the request, the dataset, the task - in the thread where the people concerned already are, instead of a log or a report nobody opens.

So the ordinary things work on it. Somebody replies underneath. Somebody quotes the part they disagree with. Somebody reacts and the matter is settled. The agent's finding and the human argument about it are one thread, in the place they are both about.

And it is why the record survives being read a year later: what the agent found, what people said about it, and what was decided are one sequence, not three systems to correlate.

And it does not require people to be in the platform

A comment leaves the screen. When somebody is mentioned, the comment reaches them in Slack, Telegram or email - the message itself, not a "you have a notification" stub that makes them go and look.

That matters for the people at the edge of a process: the manager who approves four things a month, the colleague in another department, the person on site. They do not have to adopt a new place to be part of the discussion. And in the applications that are wired for it, a reply in their channel comes back as a comment on the record, so the thread stays in one piece even though half of it was written from a phone in a chat app.

Working on the same thing at the same time

Discussion is one half of collaboration. The other is editing something together, and it is here too.

In a spreadsheet, several people work in the same document at once: changes appear as they are made, you see who is where, and nothing silently overwrites what somebody else just typed. There is no file to send around, no copy to reconcile, and no question about which version the meeting is looking at.

That is not the same mechanism as a comment thread, and it should not be: editing a document together and discussing it are different problems. Both are collaboration; how each is built is under the hood.

A thread is a record, not a chat

The difference shows up months later, which is when it matters.

  • Quote and reply, so a long thread stays followable instead of becoming a wall.
  • Edit history on the comments themselves: what was changed, not just the final version.
  • Attachments - the screenshot, the invoice, the document under discussion, on the thread.
  • Resolve and reopen. A thread that is finished says so; a thread that turns out not to be finished comes back instead of starting again.
  • Moderation, for the rare case that needs it.

That is what gives "why is this row 40 and not 35" an answer six months later, instead of an appeal to somebody's memory.

Reactions carry weight

Six emoji on a comment, and they carry more weight than the size of the feature suggests.

They stop the "+1" thread. Somebody proposes something and eight people agree. In a chat that is eight messages that bury the proposal. Here it is eight reactions on it, and the proposal is still the first thing you see.

They are a signal you can read at a glance. Open a thread and you can see what people thought without reading it - which comment everybody agreed with, which one raised eyebrows, which one nobody responded to at all.

They make small approvals cheap. Not every decision deserves a formal step in a process. A lot of them are "does anybody object" - and a thumbs-up from the three people who matter is that answer, recorded on the thing itself, in a second, by people who were never going to write a paragraph.

And they work on ideas. A backlog of proposals with reactions on them tells you what people actually care about, from people who would not have filled in a prioritisation survey.

The reason to name them at all: teams that have reactions use threads, and teams that do not, fall back to the chat. The friction of having to write prose to say "yes, fine" is what moves a conversation somewhere else.

What happened, separately from what was said

Discussion is one half; the other is the record of change itself.

Datasets and projects carry an activity timeline: what changed, when, by whom - as an ordinary part of the object rather than something reconstructed from a log. A number that moved has a history next to the thread about why.

For everybody

Collaboration is not an upgrade tier. It is available to all users on the platform, on all the objects above, because a discussion feature that only some people have is a discussion feature nobody uses.

Next

Under the hood - one anchor for every kind of object, and the two mechanisms behind discussing something and editing it together.