Six months after ChatGPT arrived, nearly every project manager we know has tried it, and most have an opinion. Some call it a revolution. Others call it a toy that makes things up. After a few months of careful experiments, we think both camps are partly right, so here is an early, honest look at what generative artificial intelligence (AI) does for a project manager’s week.
Where it helps
Think about a typical Monday. You have a long email thread to digest, a status update to write, and meeting notes from Friday that nobody has cleaned up. Generative AI is good at exactly this kind of work. Paste in a messy set of notes (after removing anything sensitive), and you get a clean first draft in seconds. Ask it to summarize a 40-message thread, and you get a reasonable starting point for what was decided and what’s still open.
It’s also useful for thinking. If you’re planning a data migration and want a list of risks to consider, an AI assistant will produce a long one. Most of the items you’ll already know, and a few you won’t. Treat it like a brainstorming partner who has read a lot but never worked on your project.
Rewriting is another strong suit. The same update needs to read differently for an executive sponsor, a technical lead, and a client’s finance team. AI can produce three versions of the same message in the time it takes to make coffee.
Where it doesn’t
Notice the word “draft” in everything above. These tools don’t exercise judgment. They don’t know that your sponsor dislikes surprises, or that the vendor’s lead engineer is about to leave. They can’t build trust with a stakeholder, and they can’t be held accountable. You can.
They also state wrong things with great confidence. The tools can make up facts, dates, and sources that sound entirely plausible. A status update that invents a milestone is worse than one that arrives late.
Three guardrails to set now
- Protect confidential data. Don’t paste client information into public AI tools. You often can’t know how your input will be stored or used. If a document is sensitive, don’t use it as input, or remove names, figures, and anything identifying first.
- Check every fact. Read AI output the way you’d read a new team member’s first report: carefully, with a pen. You own what goes out under your name.
- Agree on a team policy. Most organizations have no written guidance yet, so people are making private choices. Spend an hour deciding which tools are acceptable, what data stays out, and when to tell a client that AI helped with a deliverable.
Key message: “Let AI write the first draft. Keep the judgment, the relationships, and the accountability for yourself.”
If you haven’t tried these tools yet, pick one low-risk task this week, such as summarizing your own meeting notes, and compare the result with what you’d have written. You’ll learn more from that than from any headline, and you’ll have a better sense of what to put in your team’s policy.