The AI chat in the task details answers questions about the open task and its project, and carries out individual changes to the task at your instruction. You write in natural language, and the AI responds in the chat and links to tasks and people. This article explains how to open the chat, what you can ask it, which changes it makes and what it does not do.
Please note
The AI chat is one of the AI features. It is available when AI is enabled in the workspace and you are allowed to see AI features. Each message uses credits; follow-up questions from the AI use very few credits. See How AI credits work. The AI can make mistakes. Check its answers and changes.
Note
This chat belongs to the respective task. A universal AI chat that answers questions and handles requests across projects is in preparation and will be available soon (see “Coming soon” in How smenso AI works).
Open the AI chat
- Open a task.
- At the top of the task dialog, click
AI chat. The tooltip saysOpen AI chat.
The chat opens to the right of the task with the heading “How can I help?”. The note at the top says: “This chat is not saved and is discarded for privacy reasons when you leave the task.”
At the bottom, you will find the input field Ask smenso…. Above it, Refers to shows which task is currently open. You can send your message with Enter or the arrow. Use Attach to add files, usually PDFs and images that the AI can analyse. Multi-line inputs retain their line breaks.
Quick actions
As long as you have not written anything, the chat offers four quick actions that you can click. Clicking one sends the respective instruction for the open task:
-
Write description: The AI writes a structured description based on the task data and replaces the existing description with it. In the test, the resulting text included sections such as “Goal” and “Key technical implementation areas”. The chat summarises what it wrote. -
Improve title: The AI corrects the title with regard to spelling, grammar and readability without changing its content, tone or length. If the title is already fine, it remains unchanged. -
Summarise task: The AI writes a summary of the task in the chat, including an assessment, status, risks and next steps. It also appears as a card at the top of the task. See AI summary. -
Create subtasks: The AI is supposed to suggest suitable subtasks. In practice, this did not always work in the test. If the task already had subtasks, the chat replied that it could not expand them. In this case, useSuggestin the Subtasks section for suggestions. See AI subtasks.
The quick actions disappear after the first message. Formulate further requests in the input field.
Ask questions
The chat knows the open task, its project and the other tasks in the project. It answers questions on three levels:
- About the open task: effort, due date, assignees, subtask status, for example “How much effort is planned?” or “Is the task overdue?”
- About other tasks in the project: “What is the status of ‘Create schedule’?” or “Compare this task with ‘Create schedule’.” The AI names the exact title of the task. If no such task exists, it says so.
- About the entire project: “What is overdue in the project?”, “What do I have due next week in the project?”, “How is the project progressing?” or “What has happened in the project since Monday?”
Dependencies
The chat explains what the task depends on, what is currently blocking it and who is waiting for it to be completed, for example in response to “What is this task currently dependent on?”. It names the affected tasks and their status. If there are no dependencies, it says so.
Similar and duplicate tasks
With questions such as “Does this task already exist?” or “Are there similar tasks in the project, including among the completed ones?”, the chat searches for duplicates in the project. It only names tasks that are genuinely similar in content, not those that merely have one word in common. The task’s own subtasks do not count as duplicates. It names the search terms used in the response.
Change history
The chat answers questions such as “What has changed in the task this week and who made the changes?” using the change history: It states what was changed, the old and new values, when it happened and who made the change. It answers questions about the change history and what has happened over the past 90 days. It answers forward-looking questions, such as requests for a project overview, for the next 30 days. If you ask about a longer period, the chat points out this limit.
Workload
With questions such as “Who has the most outstanding work in the project?” or “How busy is Homer?”, the chat provides figures for each person: open tasks, tasks in progress, overdue and upcoming tasks, and planned effort. If the person named does not exist in the project, it says so.
Answers and links
Tasks and projects are linked in the answers and open in a new tab. If you hover over a person’s name, their profile card appears. The response states which steps the AI carried out, for example “Search tasks in project” or “Check dependencies”. The chat responds in the language you write in.
Changes to the task
At your instruction, the chat changes individual task values, for example:
- Priority: “Set the priority to high.”
- Due date and effort: “Set the due date to next Friday.” or “Increase the planned effort by 3 hours.” The chat can also estimate the effort (“Estimate the effort and enter it.”) and briefly explains how it arrived at the value.
- Title and description: rephrase, make friendlier or shorter, correct spelling, expand or translate. With “Correct the spelling”, the content remains the same. With “Expand the description by one paragraph and then translate everything into French”, it expands it first and then translates it.
- Translations: currently available in 46 languages. If a language is not available, the chat tells you.
- Subtasks: suggest them. See AI subtasks.
The change is applied directly to the task. You see it immediately in the fields of the task dialog and can continue editing it as usual. You can make several requests in one message, for example “Shorten the title and set the priority to high.” The response then addresses each part. In the example, the chat set the priority to High, the due date to 30/10/2026 and the planned effort to 8 hours.
These changes appear in the task activity history under your name, as if you had made them yourself. They are not marked as AI changes there.
What the chat does not change
If the chat cannot carry out a change, it politely explains why and leaves the task unchanged at that point. This applies, for example, to these requests:
- Status and assignment: The chat does not carry out “Set the status to completed.” or “Assign the task to Homer.” In the test, the response was approximately: “The task could not be assigned to Homer because this change is currently not supported.”
- Incompatible values: for example, a start date after the due date.
- Costs without a basis: for example, “Increase the costs by 10%.” for a task without any costs.
- Data from other tasks: The chat reads other tasks but does not change them.
- Unsupported languages for translations.
It still carries out the actionable parts of the same message, such as changing the priority. After a rejected change, feel free to check the task to make sure everything remains as it was.
Follow-up questions for unclear requests
If a request is ambiguous, the chat asks a follow-up question instead of guessing. The follow-up question appears in a window above the input field, with options to choose from. Click the appropriate option or select Other… and enter your answer yourself. This happens, for example,
- when a translation is possible in several variants (“Should I translate the title into Simplified or Traditional Chinese?”),
- when a period or scope is not specified (“This task” or “The entire project”),
- when several people match a name: If, for example, there are two people with the first name Homer in the project, the chat asks who you mean and names both people. If two people have the same display name, it distinguishes them by their email address. After you make your selection, the response refers only to that person.
- when several tasks match your description, for example several tasks with “Planning” in the title. The options show the task’s code and title.
- when a name cannot be found but there is a similar one: The chat may ask, for example, whether you mean the person “Marco” when you entered “Marko”.
Unambiguous names do not result in a follow-up question. If there is no matching or similar name, the chat says that it could not find anyone with that name. It only recognises short names with fewer than three letters when there is an exact match: “Al” therefore means Al, not Alfred. If you make several requests in one message, the chat completes the unambiguous parts immediately and only asks about the unresolved part. Use the × symbol to close the follow-up question without sending anything. If you write a new message instead, the chat discards the follow-up question. If the chat has several questions, it displays them one after another, and you can jump back and forth between them.
Related articles
FAQ
Is my chat history saved?
No. The chat is not saved and is discarded for privacy reasons when you leave the task. Changes made to the task by the chat are of course retained.
Does the chat change anything without my instruction?
No. The chat only becomes active in response to your message.
Can you see in the activity history that the AI changed something?
No. Changes made by the chat at your instruction are listed under your name in the activity history.
What is the difference between this and the AI actions in comments?
The chat is an open conversation about the task and the project. The AI actions in the Comments tab specifically analyse the discussion. See AI comments.
Is there a chat for the entire project or workspace?
Although the chat in the task details can also read other tasks in the project, it remains tied to the open task. A universal AI chat is in preparation and will be available soon.
Comments
0 comments
Please sign in to leave a comment.