When AI Drafts the Status Report, the Project Manager Role Moves to Three Places
When AI drafts the status report, summarises the meetings and chases the action owners, the project manager role moves to three places: deciding what the tools may do without you, checking what they produce, and finding what each task was quietly doing before you hand it over. The admin is going quickly, now that assistants sit inside Word, Google Docs, Teams and the slide tools, and always-on agents keep working between conversations. The hours it frees won’t land on the judgement work unless you put them there.
The Admin Is Leaving Through the Tools You Already Use
None of this needs a new platform. Copilot summarises a Teams meeting and lists the actions. Gemini and ChatGPT draft inside Google Docs and Word, and all four major assistants build a deck from your documents. On 29 September OpenAI launched always-on agents with their own cloud computer that keep working after you close your laptop.
Each of these takes a piece of the coordination work: the formatting, the chasing, the first draft. The agent governance charter covers the agents your team runs. This piece covers your own desk, and each job below maps to a part of the charter. Two of them repeat it in miniature. The third asks something the charter doesn’t.
Job One: Decide What the Tools Do Without You
Every assistant you use makes small decisions on your behalf: which actions go into the summary, how the update is worded, which owner gets the chaser. Deciding which it can make alone, and which come back to you, is the first job. It is your own line in the charter’s register.
In practice that is a short list per tool. Meeting summaries can go to the team unread. Anything sent to a sponsor or supplier comes to you first. Nothing changes a date in the plan. The Outcome Brief and the Charter Check cover how to write those limits for a single task, and the delegation register shows how to track them.
Job Two: Check What the Tools Produce
A drafted update reads well whether or not it’s right, and the more the tools produce, the less anyone reads closely. The second job is the charter’s spot-check at personal scale, on a schedule, before anything looks wrong.
Once a week, pick three things the tools produced, such as a meeting summary, an action list and a drafted update, and compare each with its source: the recording, the tracker, the plan. When the same kind of error turns up twice, change the instruction or take that task back. The QA checklist sets out two gates for it, and the model test covers a new model.
Job Three: Find What Each Task Was Quietly Doing
The status report was also how you noticed a sponsor going quiet. The action chaser was also how you heard that a team was stuck. Nobody listed those as tasks, so they leave with the task. The charter’s questions ask what a wrong output would cost. They don’t ask what the task did besides the task, and this job closes that gap. What goes missing is the people work no tool settles: two teams that each think the other is the blocker, the supplier who agreed a date on the call and hasn’t confirmed it in writing, the sponsor who nodded in steerco and asked a different question in the corridor.
Before you hand a task over, find out what else it was doing. Memory won’t tell you, so read the last four weeks of the emails, chat threads and meeting notes around it. Ask your AI tool to list everything in them that a task list wouldn’t record, with a dated example for each. Strip the names out first and check your organisation’s policy before pasting anything about real people. Then ask the people who rely on the task whether the list is right.
For each item, name where it lives afterwards: you, a colleague, or nobody. Anything that lands on nobody means the task isn’t ready to go, or needs a deliberate replacement, such as a fortnightly call with the sponsor. Then write the rule for taking the task back before you see how it goes. For example: if the sponsor hasn’t contacted me unprompted in three weeks, I take the report back. A rule written after the result gets bent to fit it.
Move the Hours on Purpose
When a tool takes a task off you, decide where that hour goes before the week fills it. Four lines in your Monday plan are enough, and they work as your personal entry in the charter’s register (these are plain fill-in fields):
Handed over: the task a tool now does, and the limit you set on it What else it did: the functions from your list, and who covers each Hour goes to: one named item under job one, two or three Checked: what you compared last week, and what you found
Run it for a month. If “Hour goes to” keeps reading “email”, the time has been freed but not yet moved.
Frequently Asked Questions
What does a project manager do when AI handles the status report? Decide what the AI tools may do without review, check what they produce against the source each week, and find what each handed-over task was also doing, such as noticing a quiet sponsor, so that work still has an owner.
Will AI replace project managers or project coordinators? AI is taking over the coordination mechanics of the role: drafting, summarising and chasing. Gartner predicted in September that by 2029, 30% of employees laid off because AI replaced them will need to be rehired, often at significantly higher cost. That covers all roles, not coordinators specifically, but it points at cuts made from task lists alone.
How do I find out what a task does beyond the task? Read four weeks of the emails, chat and meeting notes around it, ask an AI tool to list what a task list wouldn’t record, and check the list with the people who rely on the task. Then name who covers each item.
Which task have you already handed to an AI tool, and what else was it quietly doing?
Which task have you already handed to an AI tool, and what else was it quietly doing?
Yes, AI helped me to write this :)
Unsubscribe
