AI Tools for Project Managers
I was skeptical that an AI could help me run a project. A year later it drafts half my paperwork - and I trust it less than ever, in the best possible way.
When the first wave of AI assistants arrived, I did what a lot of project managers did: I rolled my eyes and went back to my status report. I had seen "productivity revolutions" before, and most of them were a new dashboard nobody filled in. I run delivery projects - real deadlines, real stakeholders, real consequences when something slips. The last thing I needed was a confident robot inventing my timeline.
I was wrong about the eye-rolling and, it turns out, right about the confident robot. Both lessons changed how I work.
The afternoon I gave in
It started, as these things do, with a Friday I did not want. I had four hours of meeting notes to turn into summaries, a status report due to a sponsor, and a difficult email to a client about a vendor delay. I was tired and the writing was the slow kind - not hard to think, just tedious to produce. On a whim I pasted my meeting scribbles into an AI assistant and asked for a summary with action items.
It came back in fifteen seconds. It was eighty percent right and, crucially, it was structured - owners, actions, dates, the open questions flagged. The twenty percent that was wrong was instructive: it had assigned an action to someone who was not even in the meeting. I fixed it in a minute. The whole task that usually ate my Friday afternoon took twenty minutes.
That was the hook. But the misassigned action item was the real lesson, and it has stayed with me longer than the time savings.
What it is actually good at
Once I got past the novelty, a pattern emerged. AI is brilliant at the work that is language and structure - and a project manager's week is full of that. Drafting a charter from a few sentences of context. Breaking a vague goal into a task list I can argue with. Brainstorming risks until it surfaces the two I would have missed. Turning a mess of standup notes into a clean record. Re-pitching the same status update for a sponsor who wants risks, a team that wants detail, and a steering committee that wants one paragraph.
The re-pitching trick alone earns its keep. I write an update once, then ask for "the three-sentence version for an executive, focused on decisions needed." Same facts, different frame, thirty seconds. I used to do that by hand, badly, at the end of a long day.
It is also, unexpectedly, a good sparring partner. Before a steering review I ask it to play a skeptical committee member and throw the five hardest questions at my plan. It finds the soft assumptions I had stopped seeing. I walk in prepared instead of ambushed.
What it is dangerously bad at
Here is where my old skepticism turned out to be a feature, not a bug. AI writes with total confidence whether or not it is correct. It will invent a milestone date that sounds plausible. It will assign an action to the wrong person. It will smooth a genuine blocker into a reassuring sentence that hides the problem. And it will agree with my plan enthusiastically even when my plan is bad.
None of that is a reason to stop using it. It is a reason to use it like a draft. I now have one unbreakable rule: nothing the AI produces leaves my hands until I have checked every fact, name, number, and date, and read the whole thing as if I wrote it. Because to my stakeholders, I did. "The AI wrote it" is not a sentence I ever want to say to a sponsor.
The line I will not cross
There is a second rule, and it is about what I refuse to hand over. AI can draft the difficult client email. It cannot know that this particular client needs to feel consulted before any change, or that my sponsor distrusts anything that sounds too optimistic. It can propose a schedule. It cannot know that one engineer is quietly drowning and the "available" slot on the plan is a fiction. It can summarize a meeting. It cannot own the decision that meeting reached.
So I delegate the production and I keep the judgment. AI does the typing; I do the deciding, the listening, and the answering-for-it. That division has become the spine of how I work.
The part nobody warned me about
The confidentiality question crept up on me. The first time I started to paste a client's contract terms in to "just clean up the summary," I stopped cold. That data does not belong on someone else's servers. Now I anonymize by reflex - Client A, Vendor X, placeholder numbers - and I know which tool my company has approved and what its rules are. It is a small discipline that prevents a large mistake.
Where it leaves me
A year in, I draft maybe half my project paperwork with AI, and I trust the tool less than I did on that first Friday - which is exactly right. I trust it the way I trust a fast, capable, slightly unreliable new assistant: grateful for the speed, never asleep at the review. The hours it gives back do not vanish into more paperwork. They go to the work that made me a project manager in the first place - thinking ahead, protecting the team, reading the room, and making the calls no tool will ever make for me.
If you are still rolling your eyes, I understand. I was you. Try it on one tedious Friday task. Then check its work like your name is on it. Because it is.
No comments:
Post a Comment