Agentic Project Management

A Project Management Problem

Recom's project management system had 75+ open project plans, almost none of which were current. Department heads were staring at tasks from 2020 that had never been completed. Untracked subtasks were multiplying, creating the illusion of progress without substance. Teams ran the same unresolved cycle week after week.

The system wasn't failing because people were lazy. It was failing because it was designed incorrectly. Too many tasks and too much structure.
A Fresh Framework

I brought the team a redesign built around one core shift: stop managing tasks, start managing outcomes.

The new structure had three components. First, a singular project recap task per brand; one place where the full picture lived, updated on a defined cadence, visible to every stakeholder. Second, department-level tasks to replace the tangle of subtasks. Each department's point of contact owns one update that synthesizes everything happening in their area. Third, a hard rule on when subtasks are warranted: only when a deadline-driven item genuinely needs its own tracking. Everything else stays in the department update.


The Agent Builds

Once the framework was stable, I scoped three AI agents to run the operational layer that humans had been doing manually.

AGENT A

Recap compiler. Pulls department-level task updates from across the client’s project plan and synthesizes them into a single partner-facing recap, posted to the project recap task. A human reviews for accuracy and relevance before it goes anywhere.

AGENT B

Meeting notes creator. Once a human approves the weekly recap, this agent formats it into structured meeting notes following a defined template. Deliberately kept separate from Agent A because the team wanted control over when notes get created, and collapsing both into one agent was creating more anxiety than efficiency. 

AGENT C

Department task updater. An AI note-taker runs during client calls. After post-meeting notes are finalized and human-reviewed, this agent reads them and posts relevant updates to the appropriate department tasks.

Next
Next

Signal