One of our AI Teammates is supposed to build a project status update based on tasks that were recently modified. The project in question contains hundreds of tasks, and based on our extensive testing, we realized that this behavior was not consistent enough. Some tasks were completely ignored for example, or old tasks were seen as “updated”.
We’ve been wondering if the problem is simply a context that is overloaded, so we decided to test a new approach.
The idea is to create an automation that will place modified tasks into a separate project, and have the teammate build a status update for that specific project. Once the update is done, you can empty the project so it can fill up again.
It definitely feels like a workaround, and hopefully this can be fixed in the future. But just in case, if you’re stuck with a similar problem, this might be the solution you’re looking for.
We are [i.DO](http://ido-clarity.com), an Asana Solutions Partner, and we document our learnings using AI Teammates every day!
I did not. Those bug cases are a nightmare to deal with. The support would ask a ton of questions, and then it would take weeks for them to simply say they are “working on it”, when in fact I believe they just can’t replicate… We had exactly the same issue with the “AI Smart Status” and they never managed to fix it.
I am hoping this is larger problem that’s being looked at.
Thanks for sharing this workaround! That’s a really interesting approach, and I can see it being useful for larger projects where the AI teammate has a lot of context to work through.
Out of curiosity, have you also tried refining the AI teammate’s instructions or saving guidance to its memory? I’d be interested to know whether that had any impact on the consistency.
One thing that may help is using more specific rule triggers than a broad ‘task updated’ condition, like completion changes, due date changes, new comments, or updates to a key custom field.