# Collaboration

> The conversation about work, attached to the work: comments, mentions and reactions on the record, the dataset, the cell range or the block, not in a chat that lost the context on the way

Source: https://aipril.dev/docs/collab

## Where decisions currently go to die

Somebody notices the number is wrong. They screenshot it into a chat.

Four people reply. Two of them are talking about a different month. Somebody explains why it is
actually correct. Somebody else asks a question that gets buried. A decision is reached, roughly,
by the fifteenth message, and nobody writes it down anywhere because everybody was there.

Six weeks later the same number comes up. Nobody can find the thread - it was in a channel that has
had four thousand messages since, and the screenshot does not say which version it was. So the
conversation happens again, and the person who explained it the first time is on holiday.

**None of that is a communication problem.** The tools worked exactly as designed. The problem is
that the discussion was never connected to the thing it was about.

## What that costs

- **Context is reconstructed, repeatedly.** Every conversation starts by establishing what everybody
  is looking at, and half of them establish it wrong.
- **Decisions are not findable.** "We decided this in March" is a claim, not a record, and the only
  way to check is to ask somebody who remembers.
- **The people who should see it do not.** A thread reaches whoever is in that channel, which is
  never exactly the set of people concerned.
- **And the work does not know.** The task moves on, the report is rebuilt, the plan is approved -
  none of them carrying the reasoning that produced them.

## What is different here

**The thread is on the thing.** A comment attaches to what it is about: a record, a dataset, a range
of cells in a spreadsheet, a project, a checklist item, a block on a page being designed. Open the
thing next year and the conversation is there, with what it was about, because it never left.

**And it reaches people where they are.** Mention somebody and it arrives in Slack, Telegram or
email, with an inbox of what is waiting for them, instead of a hope that they open the right screen.

**And your AI employees are in it.** An agent that checked something, gathered something or noticed
something writes what it found as a comment on the record, in the thread where the people concerned
already are, not in a log somebody would have to go and read. A person replies underneath. See
[AI organisation](https://aipril.dev/docs/ai-organisation.md).

**Working on the same thing at the same time is the other half.** Discussion is one mechanism;
editing a document together is the other, and both live here. Two people in the same
[spreadsheet](https://aipril.dev/docs/spreadsheets.md) see each other's changes as they are made - no file, no copy, no
question about which version the meeting is looking at.

That is the whole idea, and it is deliberately small. Collaboration here closes the gap between
talking about the work and the work itself.

## Where to go next

- [What you can do](https://aipril.dev/docs/collab/what-you-can-do.md) - threads, mentions, reactions, agents in the
  thread, co-editing, and everywhere they attach.
- [Under the hood](https://aipril.dev/docs/collab/under-the-hood.md) - one anchor for every kind of object, the two
  realtime mechanisms behind discussion and co-editing, and how access is decided.