Connecting your tools

Connect Asana to an AI agent

A tracking tool does not say what is done: it says what someone took the time to declare done. The whole gap is there, and it grows with the workload — the week you work most is the week you update the board least. Connecting an agent to Asana changes something simple and unusual: whoever carries out the task is also whoever ticks it, at the moment they finish it. The board stops being a report written afterwards and becomes the trace of what took place.

1. The gap between work and its declaration

All tracking tools share the same blind spot, and Asana is no exception: they measure declarations. A ticked task means a person opened the tool and clicked — not that the work is finished, nor that it is well done.

This gap has three effects, and the third is the costliest:

The board lags behind reality. What is done on Tuesday is ticked on Thursday, or the following Monday. Over a week that changes nothing; over a quarter it is enough to make a report false.
Updating becomes work in itself. A weekly meeting whose real purpose is to keep the tool up to date is a meeting that costs everyone's time to produce a display.
You end up not using it any more. A board you know to be false is no longer consulted. It stays open, it no longer decides anything, and the real priorities go back to living in messages and in people's heads.

2. What the agent does in Asana

Asana's official connector covers projects and tasks, with what is needed to track their progress.

View projects and tasks. Knowing what is in progress, what is late, who carries what. That is what lets an agent know whether it should get to work — or whether someone is already on it.
Create a task. What comes up along the way and would otherwise stay in a discussion thread: a correction spotted, a follow-up to make, a missing item to ask for.
Assign a task. To a person on your team, with a deadline. A task with no owner and no date is not a task, it is an intention.
Track progress. Moving forward what has moved forward, flagging what is blocked, updating the status rather than letting it go stale.

3. The difference: the one doing the work keeps the board

This is the point that really sets this connector apart from the others, and it is worth insisting on because it is easy to miss.

In the usual use of Asana, the person doing the work and the person keeping the board are the same — which is precisely what doesn't work, since the second task is always postponed in favour of the first. When an agent carries out the task, the situation changes in nature: it has no reason to postpone the declaration, and it makes it at the exact moment the work ends.

This gives Asana a function it did not have: that of a log of what actually happened. A task “publish this week's article” is no longer ticked out of optimism on Friday; it is ticked when the article is online, with the trace of what was done.

Two practical consequences:

The board becomes consultable again. If it says a thing is done, it is — which, in a tracking tool, is less common than one would like.
Humans and agents share the same list. What is delegated and what is not appear in the same place, with the same deadlines. It is the simplest way to see what your agents are doing without learning a new screen.

4. What it does not do

It does not tick what it has not done. A task carried by a person stays with that person: the agent can ask where it stands, it does not declare on their behalf. A board kept by an agent that assumes is worth less than a board that is late.
It does not reassign other people's work. Moving a task from one person to another is a management decision, not a tidying operation.
It does not change deadlines to avoid a delay. A postponed delay is a hidden delay; it must stay visible, including when it is the agent itself that is late.
It does not restructure your projects. Your sections, your tags and your templates remain your organisation.

5. Connecting, and naming the project

Connecting is an account authorisation: you log in at Asana, you accept the permissions shown. No application to create, no key.

The scope is named on two levels:

The workspace — the Asana organisation concerned.
The project — the precise list the agent works on. The other projects in the workspace stay invisible to it.

Limit access to the projects your agents must work on. This advice seems obvious and is regularly neglected, because an Asana workspace almost always contains projects that have no business being within reach of a program: a recruitment under way, a reorganisation, a confidential file. Naming the project settles the question once.

It is also the right way to start: a single project, the one whose tracking goes stale fastest, then widen from there.

6. What it changes over time

A tracking tool is judged on its reliability after six months, not on its comfort in the first week.

Week 1. Writes go through validation and you set the vocabulary: what deserves a task, what does not, at what moment a task is finished at your company. It is subtler than it looks and it is only settled once.
Month 3. What has been done is ticked the same day. The review meeting stops being an update session and becomes a decision session again — which is its purpose.
Year 2. The history is complete, and therefore usable: how long this kind of task really takes, which one always slips, which one never served any purpose. That reading is impossible on a board kept intermittently.

No quantified productivity gain is promised: it depends on what you delegate. What is guaranteed is that the board tells the truth when you open it.

7. Frequently asked questions

Asana or Notion to steer agents?

Both connect, and they don't answer the same question. Notion holds documents and databases: what you know. Asana holds assigned tasks with deadlines: what must be done, and by when. Many companies use both, and that is coherent as long as each keeps its role.

Will my team see that the task came from an agent?

Actions appear under the name of the Asana account you connected — all the more reason to dedicate one to it rather than use a personal account. It is also what makes the board readable: you can tell at a glance what comes from the team and what comes from the agents.

Can it work the other way: can an Asana task trigger an agent?

It is the most natural use once the connector is in place: the agent looks at the project, sees what falls to it and gets to work. What triggers its work is what you have defined in nullbot; Asana serves as a shared list, not as an engine.

What happens if I disconnect?

Access is cut immediately, and the authorisation can also be revoked from Asana. The tasks created, the assignments and the history stay in your projects: nothing is withdrawn. What stops is the updating.