Hello Asana-verse !
Just saw today that there were some new Asana AI default custom fields to add such as generating an AI Summary for unblocked tasks, that provides more context to the assignee of the unblocked task without digging into prior task(s) !
Loved this idea so went into my sandbox to test out and here were some findings and limitations, including uncertainty about Asana Ai’s ability to pull from blocking task and not context from the blocked task, and whether there is API limits when there are multiple dependencies.
Questions for the community:
- Comment inclusion: In my testing, task comments from the blocking task never appeared in the AI summary output — only the task description was used. Has anyone found a way to include comment content in the summary, or is this a known platform limitation?
- Multi-dependency behaviour: When a task has 2+ blocking dependencies and all are completed, the summary only restates the blocking task names rather than synthesizing their descriptions. It seems like full descriptions are only passed for single-dependency scenarios. Has anyone experienced this differently, or found a workaround?
- Source task context: Is the AI reading from the blocking task’s description, the unblocked task’s description, or both? My testing suggests it reads the unblocked task’s context — which means AI guidance instructions to “read from the blocking task only” cause the rule to return empty output. This contradicts #1 and surfaced through multiple iterations of AI guidance adjustments.
Would love to hear how others are using this rule in production and whether these are known limitations or configuration issues on my end. Thanks!
Hi @Sarah_Yeung, loved this!
I tested this with one blocking task that had a comment and two dependent tasks. The AI generated a summary for both dependent tasks, but included the comment only where it seemed relevant to completing that specific task. This suggests the AI may be selectively using relevant information from the blocking task rather than including all comments in every summary.
This is the comment AI sent for Dependent Task A (comment of precedent task is relevant)
Comment for Dependent Task B
So if the same blocking task has two dependent tasks:
- Dependent Task A: The comment is relevant → the AI includes it.
- Dependent Task B: The comment isn’t relevant → the AI may leave it out.
That actually suggests the AI is selecting relevant context for each dependent task, rather than blindly copying all comments from the blocking task into every summary.
Hi @Irish_Makimura thanks for the share !
That helps with AI pulling relevant comment details into the summary - I’ll test this again because this was not the case for me. But to clarify, for #2, I was asking about the inverse of your example instead of 2 dependent tasks, I have 1 dependent task and 2 blocking tasks. When all blocked tasks are complete, are the full description of each blocking task passed to the AI ? My testing showed only the blocking task names were passed to the AI
Could you also affirm #3 for where the AI pulls its details from? From the blocked task or the blocking tasks ?
Thanks again !
Hi @Sarah_Yeung,
Thanks for clarifying 
I tested the 1 dependent task + 2 blocking tasks scenario.
In my test, the AI was able to pull information from both blocking task descriptions when both precedent tasks contained meaningful descriptions. So I did not see only the task names being passed in that specific setup.
For #3, based on my testing, the summary appears to be generated primarily from the blocking tasks (the precedent tasks) rather than from the blocked task itself. The blocked task seems to provide the relevant context, but the details in the summary came from the blocking task’s descriptions and, when relevant, comments.
One thing you could try is making the prompt a bit more explicit about combining information from all completed blocking tasks, for example:
“Summarize the work completed in all blocking tasks. Include relevant details from each blocking task’s description and comments, and combine them into a single summary for this task.”
That may help in cases where the AI is being too selective about which blocking-task details to include. Hope this helps.
Hi @Irish_Makimura ,
I’m so sorry, I just realized that maybe we’re talking about two separate types of Ai rules. In your screenshot, may I confirm if you’re referencing an AI generated comment on the blocked task, referencing work completed on the prior blocking tasks?
If so, I was talking about the new AI custom fields that Asana released which is in the task description panel and my goal was to get a concise enough summary since the it expands the custom field and can be too text heavy. Here are my examples:
The second one here is where the Ai summary generated only the names of the blocking tasks when there were two blocking tasks:
So I was wondering if that was a limitation for this AI custom field and if anyone else has encountered the same?
Apologies for the back and forth!
I appreicate the insight into task comment summaries though (: I was just testing out a new feature to see if we can minimize the comment notifications, especially when there are a lot of back and forth cross teams, and consider whether a specified field can be an easier reference source for the user vs scrolling through comments !
I am so sorry @Sarah_Yeung,
You are right! I totally thought it was the AI studio Rule “Summarize blocking task”
I tested the AI studio field, and this is the result I got; it gave me a summary of both blocking tasks, but I agree it’s too heavy.
I tried updating the prompt in Rule to request a more concise summary while retaining the key information and comments, and the results were much better.
I think this is one of those AI-generated fields where tweaking the prompt can really improve the output. I’ll also share this feedback internally. Thanks for testing this out, Sarah! I really enjoyed the discussion and the different scenarios you brought up!
Thank you @Irish_Makimura ! That’s really helpful. I’ll continue tweaking my AI guidance. My handful of tests show that it either is too long, or the AI hallucinated details in the AI summary, or too short (ex. when quoting only the title of the blocked tasks vs summary of what was fixed). Ideally the AI could pull relevant fixes from the blocked task(s) so they don’t have to dig into the previous task to check what was done. In your second image it seems your AI summary gave context to each blocking task vs my second image citing only the task names.
I’ll keep iterating, but good to note this is possible ! (: Thanks again!
