Asana and AI agents

When work runs through several people, it isn't lost inside the tasks — it's lost in the joins between them. A blocker clears and nobody tells the person who was waiting. A date moves by three days and nobody notices the four tasks standing behind it. The agent works in the projects you open to it and watches exactly those joins: what's blocked and on whom, what was promised for a date, and what has quietly slipped.

What an agent does in Asana

  • Works in the projects you open to it, and stays out of the rest.
  • Keeps track of what's blocked and on which person, not just on which task.
  • Tells the assignee their task is unblocked instead of waiting for them to notice.
  • Works out what a moved date drags behind it and shows the whole chain, not the one task.
  • Spots a handoff nobody picked up and raises it before the week is gone.

Processes that run themselves

A playbook isn't a button — it's a process with a condition that starts it. In Asana that condition usually comes from the link between tasks: a blocker cleared, a date moved, an assignee changed.

The blocker clears

The task another one was waiting on is marked complete

  1. Check that the thing being waited for actually exists: the file, the answer, the decision
  2. Tell the assignee of the waiting task that they can start
  3. Come to you if the blocker was closed with nothing behind it

A date moves

A due date on a task is moved

  1. Find everything that was standing behind that date
  2. Assemble the chain: who promised what to whom further along it
  3. Show how far the slip reaches and where it breaks a promise made outside the team
  4. Bring you the list — it doesn't move other people's dates

A handoff nobody picked up

A task is reassigned and the new assignee doesn't pick it up

  1. Ask in the task whether the new assignee has taken it on
  2. Wait the interval you set, and not chase more often than that
  3. Raise it with you, with the previous assignee and the date of the handoff

Where the agent stops

  • It doesn't set priorities: what matters more is your call — it shows the consequences, it doesn't choose.
  • It doesn't reassign other people's tasks or move their dates: that's a question first, not an action.
  • A task it can't state unambiguously comes back as a question rather than a guess.
  • Asana is only as true as what people record in it, and when the agent is inferring a link rather than reading one, it says so.
Connect AsanaSetup is a conversation, 15–30 minutes

FAQ

Will it message people on its own?

In the tasks where you've allowed it, yes — as a comment, and only about something concrete: unblocked, not picked up, date moved. You set how often it may chase. A bot that posts every day stops being read in its second week.

Do dependencies have to be set up properly in Asana?

It helps, but it isn't required. Where dependencies exist, the agent follows them. Where they don't, it works from dates, projects and sections, and says plainly that it is inferring the link rather than reading it.

Will it change due dates?

Its own, the ones you handed it, yes. Other people's, no. Someone else's date is a promise they made to a third party, and that isn't moved quietly.

Do we need an Asana admin to set this up?

For organisation projects, yes — an admin connects the app. After that the setup is a conversation: which projects, how long a handoff may sit unanswered, what counts as a handoff at all.

Where this shows up

Walkthroughs where Asana carries part of the process.