Task Assistant← Back to demo
Sample walkthroughTask Assistant

Sample walkthrough — Task Assistant

This public walkthrough uses the fictional Northstar board. It has no Asana connection, external AI service, or task-system account; confirmed changes stay in this browser. Start from the original fixtures: choose Reset sample, review the replacement preview, and confirm it. Reset replaces local browser edits.

  1. In Workspace, choose What is blocked?. The guide answers from the visible task cards and identifies NTH-108 — Instrument workspace-created event and NTH-119 — Summarize churn interviews. This is a read request: no change preview opens.
  2. Ask When is Validate admin invite flow due?. The response grounds the answer in NTH-115 and its Sep 11 sample due date. Then ask Mark that done. Because NTH-115 was the clear preceding task, the guide prepares a status proposal instead of changing it. Select Cancel. Now ask What is blocked?, followed by Mark that done. The board-wide answer expires the earlier NTH-115 context, so the guide asks which task you mean and opens no proposal.
  3. Repeat the NTH-115 due-date question, choose Clear chat, and then ask Mark that done. The workspace again asks for a task because transcript cleanup expires its hidden task context. Ask Mark Finalize onboarding checklist blocked: NTH-104 has no blocker reason, so no proposal opens and the direct Status menu disables the same transition with reason required. Both entry points use one policy.
  4. For the primary update, enter Mark Finalize onboarding checklist as done. The preview announces 1 affected task, pairs NTH-104 with Finalize onboarding checklist, and labels Before — Status: In progress and After — Status: Complete. Select Cancel once to confirm the card remains In progress; repeat the request and select Confirm change to apply the local sample update.
  5. Use the Owner menu on Review trial nurture copy and choose Jon Bell. The same preview pattern shows Priya Shah → Jon Bell before confirmation. Select Confirm change. Direct controls do not bypass the conversation safety boundary.
  6. Choose Undo last change. Undo itself opens a preview containing the NTH-112 owner difference; confirm it to restore Priya Shah. NTH-104 is still Complete from step 4. Open Reset sample and confirm the NTH-104 Complete → In progress difference. The reset creates Undo last change because it changed a visible task value.
  7. Open Reset sample again on the restored board. Now the review states that no visible task values will change. Confirm it: the guide says no new Undo step exists. Compatible saved data with title, project, due date, priority, blocker, task-code, added-record, or removed-record differences would name those changes in the same review.
  8. Choose Complete tasks in review while NTH-112 is In review. The preview lists every affected task before it can proceed. Clear chat removes only conversation messages; it does not close a preview or mutate a task; select Cancel to close the bulk preview.
  9. On NTH-104, enter Waiting on legal approval in Blocker reason. Choose Review reason and inspect None → Waiting on legal approval; cancel once, then repeat and confirm. NTH-104 remains In progress, with the reason visible as recorded context. Its Status menu now offers Blocked. Selecting Blocked opens a separate review; cancel or confirm it deliberately. A chat request such as Set blocker reason for NTH-104 to legal approval directs you to this field and prepares no change.

The request field states its 500-character, one-clear-change boundary, and the guide states that this tab retains the most recent 40 messages. These are local reviewability limits: reaching them does not modify task state, expand assistant capability, or imply an external conversation history.

The intended tradeoff is visible friction in exchange for a reviewable update. The saved-board limit is 50 tasks so this card-based recovery surface remains inspectable; a larger workspace would need search, pagination, and a different review design. Whether real users understand and value these tradeoffs is unanswered; the proposed study compares them with each participant’s current workflow before broader language support, integrations, or automation.