AI teammate for analysing the communication gap

I’m exploring a use case for an AI Teammate at the Portfolio level and would appreciate some guidance on whether this is achievable.

Use Case

I would like the AI Teammate to continuously monitor every project within a portfolio, and then scan every task inside those projects for communication activity.

The AI should:

  • Review all task comments across every project in the portfolio.

  • Identify comments where no reply or acknowledgement has been made for more than 24 hours.

  • Treat these as potentially unattended conversations that may require follow-up.

Expected Output

Every 4 hours, the AI should generate a consolidated summary at the Portfolio level containing:

  • Project Name

  • Task Name

  • A brief summary of the pending communication

  • Who posted the unanswered comment (if possible)

  • How long the comment has remained unanswered

  • Any potential blocker or action required

For example:

Project: Website Redesign
Task: Homepage UI Review
Issue: Comment from John requesting design approval has not received a response for 28 hours.

The goal is to provide project managers with a proactive communication health report so that no requests, questions, or blockers are missed across the portfolio.

What I’ve Tried

The closest approach I could think of was having the AI update the Portfolio Notes every four hours with this summary. However, I haven’t been able to configure the AI Teammate to write or continuously update the Notes section as it is saying it doesn’t have editable access or can’t directly write

Has anyone implemented something similar or found a better way to surface this information at the portfolio level? Any guidance or best practices would be greatly appreciated.

Hey @Bhumika_Bihani

Short answer: not natively as one fully autonomous portfolio-level workflow today; but possible via a task assigned to an AI Teammate and the right instructions and access

A practical idea:

  • Create a task assigned to an AI Teammate (could be in any project, or not in a project)
  • In the task description, explain the mission, ask every X hours, to review the selected projects (or the same scope as the portfolio) - AIT should have it has access to all projects
  • Ask it to post a structured summary and add your criteria (“Review all tasks… no reply…”)

visual created by a AI Teammate :smiley:

Remarks:

If the AI has to scan too many tasks/comments, there is some risk that context size becomes a limitation and some items are missed. The exact limit is not known and depends on the tasks’ content but reducing scope generally improves consistency.

We use a similar technique (AI Teammate in a task) to review a large internal projects and identify recent conversations and create an update for the team (allowing users not to track all individuals tasks, but rather read a more “macro” summary).
But we recently had some issue with the size of the context CC @Bastien_Siebman

A few thoughts:

  • Have the Teammate deliver its report as a task, rather than as portfolio notes. Use a dedicated project for those report tasks.
  • Consider using a rule to create a new task assigned to the Teammate containing the instructions, every X hours (technically you’d need 6 rules daily rules, each set to a specific four-hour mark)?
  • The Teammate could then its report in a new task, or in the assigned task by replacing the original description or adding a comment.
  • I wonder if there’s a way to use Rules to create a short list of objects for the Teammate to review on each run, instead of having to read through all tasks on all projects every run. For example, when a comment is added to a task, maybe
    • Tag the Teammate and use the Teammate’s Guidance to instruct it to maintain a list of comments to check. Then on the four-hour run it can check the short-list it’s maintaining?
    • Create a new unassigned task on a placeholder Project with a link back to the comment. Then the Teammate’s guidance is to check every open task on that project, find the triggering comment, and see if somebody’s responded

Thank You @Arthur_BEGOU @Liohn_Sherer

Let me try these suggestion.

Let us know what you find

I have tried 2 ways one created a task in sperate reporting project where it is giving me exactly what was needed

The other solution which i tried is creating a gap custom field which teammate will automatically update based on duration and the will reflect at the portfolio level dashboard.

Interesting. So just a single running task that will get updated with new comments every four hours?

What’s the trigger for the Teammate, is the four-hour cycle just a line in its Guidance window or you’re using another mechanism to trigger it?

Yeah, I would also like to understand how the teammate is instructed to comment every 4 hours

@Shyamal_Yahoo @Bhumika_Bihani

Yeah, I would also like to understand how the teammate is instructed to comment every 4 hours

Actually, that’s simple, if you simply give the instruction to do something on a recurring basis ("every x days, month or any pattern), it will deliver.
The task doesn’t have to be recurring, or pushed manually by the user; or even assigned to the Teammate.

More info : Agents Playbook #3 - Get Your AI Teammate to Repeat any Action on Schedule

Yes
So trigger i have set up in the guidance only what needs to be done and fed into the memory also.

Behaviour-

You are a communication gap analyst you work on every day with 4 hrs you need to

  1. Scan portfolio-level activity across all projects, not just one project.

  2. Detect unanswered or cold communications:

  3. @mentionsmentions or requests with no response

  4. comments left without acknowledgement

  5. conversations that haven’t moved for ~24 hours or more

  6. Flag blockers / potential blockers from comments or task updates.

  7. Generate a descriptive report of:

  8. which project

  9. which task/comment

  10. who asked

  11. what is pending

  12. what needs escalation

  13. Run on a schedule, ideally:

  14. every 4 hours, or

  15. at minimum a daily scan for the last 24 hours

  16. Produce output at portfolio level as a consolidated report, not just task-level alerts.

  17. Use configurable criteria / prompt rules so the AI can classify items like:

  18. unanswered

  19. open questions

  20. blockers

  21. escalation needed

  22. on hold / at risk

use the Notes tab CME in CME instead of the custom field.

New format

  1. Project Name (Link) - Task name (Link) - gap identified.

Behavior I’ll follow

  1. scan all accessible projects inside CME

  2. post one consolidated portfolio-level update every 4 hours

  3. include an entry for each relevant project/task gap found

  4. overwrite the previous 4-hour update rather than append to it