top of page
Glossy 3D 48PEGASUS Logo.png

An AI Project Coordinator for the Work Between the Tasks

Oct 27
4 min read

Updated: Nov 12

Most projects already have somewhere to store tasks. The difficulty is keeping that task list connected to what is happening: a design is waiting for feedback, the feedback is sitting in a message, and the launch date still assumes everything is on schedule.


An AI project coordinator can bring those signals together and prepare a useful picture of the work. It can highlight missing information, explain dependencies and draft the next update. It should support the person accountable for delivery, rather than quietly choosing a new scope or making promises on the team's behalf.


Colorful sticky notes arranged across an office whiteboard
Stock photograph for illustration. Photo: Walls.io / Pexels.

Give the assistant a clear source of truth


The source of truth is the system the team agrees contains the official task status. It might be Asana, another project tool or a carefully maintained shared list. The assistant needs stable task identifiers, named owners, dates, status definitions and links to relevant documents.


Start by defining what “blocked,” “in review” and “done” actually mean. If “done” sometimes means submitted and sometimes means approved, no summary can resolve that ambiguity reliably.


Established project tools already support much of the underlying machinery. Asana documents rules that connect triggers, conditions and actions, as well as task dependencies that express which work is waiting on other work. AI can add a plain-language interpretation layer, but deterministic rules should handle fixed conditions such as an overdue date. Asana rules and task dependencies.


An illustrative example: preparing an exhibition


Imagine a company preparing for a trade exhibition. This fictional project includes a stand design, printed graphics, product samples and a staff briefing.


The print order depends on approved artwork. The artwork is marked “in review,” but the reviewer has not been assigned. Meanwhile, a project comment says the printer needs files earlier than expected. A coordination assistant could surface this gap before the weekly meeting.


Its inputs would be the project board, approved supplier dates and project-specific comments. It would not need everyone's complete inbox, unrelated client folders or private conversations.


A practical sequence looks like this:


  1. Retrieve the current project snapshot and record when it was fetched.

  2. Check for missing owners, elapsed due dates and incomplete prerequisite tasks using explicit rules.

  3. Use AI to summarize relevant comments and identify statements that might change the plan.

  4. Compare those statements with the official task fields. Treat a disagreement as something to review, not permission to overwrite a date.

  5. Produce a short exception list, with links to the underlying tasks and messages.

  6. Let the project lead approve any changes, reminders or stakeholder updates.


This sequence keeps observation separate from execution. A mention that “we might need another week” is evidence of uncertainty, not an authorized schedule change.


What a useful daily output looks like


The assistant's draft might say:


  • Artwork approval: No reviewer assigned. Printing cannot proceed until approval is recorded. Suggested next step: project lead names the reviewer.

  • Print deadline: Supplier comment requests final files by Tuesday; the task currently says Wednesday. Confirm the correct date before updating the plan.

  • Samples: Packing complete; collection remains scheduled. No action needed.

  • Decision needed today: Which artwork version should be reviewed? Two versions are attached.

  • Data freshness: Project board checked at 09:00. Supplier comments checked at 08:55.


Notice what is absent: an invented percentage-complete figure or an unexplained red rating for the entire project. A short list of actual decisions is often more useful than an impressive-looking dashboard.


A scheduled status update can use the same evidence, separating completed work, next steps, risks and decisions. Each item should have a source and a clear distinction between a recorded fact and the assistant's interpretation.


Keep approval meaningful


Approval is more than a task turning green. It needs to identify the person deciding and the exact deliverable being accepted. If artwork changes after sign-off, the old approval should not automatically apply to the replacement.


Asana's approval tasks distinguish approval, requested changes and rejection. Its documentation also notes that task permissions affect who can complete an approval. That is a reminder to verify access settings rather than assuming a named assignee is the only person capable of signing off. Asana approval guidance.


In a custom workflow, store the file version or revision with the decision. Keep the project lead responsible for deadline changes, resource allocation, supplier commitments and changes to scope. An assistant can draft options and explain knock-on effects; the accountable person selects the option.


The implementation details that matter


Reliable coordination requires more than connecting a chatbot to a board. Decide how often the system checks for updates, how it handles unavailable services and how it avoids creating duplicate tasks.


Each proposed change should reference an existing task identifier. Before applying an approved update, check that the task has not changed since the suggestion was prepared. If two people update it simultaneously, stop and show the conflict rather than overwriting newer information.


Access should follow project membership. A cross-project summary can accidentally reveal a confidential launch or customer issue to someone who cannot access the original task. Test those boundaries with different user roles.


Also set a reminder policy. Repeating the same overdue message every hour produces noise. A better design groups related issues, respects working hours and escalates only when a defined condition changes or an agreed deadline is reached.


Pilot on one project, not the whole company


Choose a live, low-risk project with a clear lead and a manageable task list. Run the assistant in read-only mode for two weeks. The lead can compare its daily exceptions with their own view before any automated reminders are introduced.


Track how many flagged issues required action, how many genuine blockers it missed and how much time the lead spent correcting updates. Look at the time between a blocker appearing and someone taking ownership. Check whether the team receives fewer fragmented status requests rather than simply more notifications.


If the pilot exposes inconsistent task definitions, fix those first. Better data may provide more value than adding more AI features. A useful starting conversation with 48PEGASUS is a walkthrough of one recent project delay: where the information existed, who needed it and which decision would have helped the work move forward.



Comments


bottom of page