A few months ago I sat in a meeting, trying to run it properly while also scribbling notes, and realized I’d missed what someone had actually agreed to do by the time it came to writing the recap.
That’s basically the whole reason I started using AI properly for this stuff. Not because it’s exciting or cutting edge, but because I was tired of trying to lead a meeting, listen properly, take notes and somehow capture every action at the same time.
AI has genuinely made my job easier. But I’ve also learnt that getting useful results from it isn’t about letting AI manage the project for you. It’s about using it to take some of the admin and information overload off your plate so you can spend more time actually managing the project.
I don’t think AI is replacing good project managers – I do think it’s changing what being a good project manager looks like.
A lot of our time disappears into things that aren’t really why we’re there: formatting updates, writing meeting notes, pulling actions out of conversations, rewriting the same information for different audiences and staring at a blank page trying to get a document started.
AI is very good at taking some of that load away.
What it isn’t good at is knowing that a stakeholder has gone unusually quiet, that an agreed date isn’t remotely realistic, that two teams clearly aren’t aligned, or that something needs escalating even though nobody has actually said so.
That’s still the job.
So I don’t really see this as AI versus project managers. I think the bigger shift will be between PMs who learn how to use AI well and those who continue doing every administrative part of the job manually.
And it gets easier. Once you’ve refined a few prompts, learnt which tools work for you and got into the habit of using them, AI stops feeling like another task and simply becomes part of how you work.
Start with the easy wins: agendas and meetings
Meeting agendas are probably the easiest place to start. I give AI the project context, what’s outstanding, who’s attending and what I need from the meeting, and ask it to draft an agenda.
It’s rarely perfect first time, but going from a blank page to 80% of a decent agenda saves a genuine chunk of time. I’ll often follow up with: “Based on the project information I’ve given you, is there anything important missing from this agenda?”
Where my organization allows it, I also use approved AI tools to capture or summarise meetings. This has probably changed my day-to-day work more than anything else.
I used to try to run the meeting and take proper notes at the same time. You can’t really do both well. Now I can listen properly, ask questions, challenge things and actually lead the meeting.
One thing I’ve learnt is to clarify vague commitments out loud. If someone says, “We should probably have that next week”, I’ll ask, “Just to confirm, are we agreeing you’ll have this completed by Thursday?” It helps everyone, but it also means the AI captures the actual commitment rather than turning a vague comment into something it wasn’t.
And when I review the notes afterwards, I don’t just look at the main talking points. I look at who spoke the least and what they said. Sometimes the quietest person in the meeting raised the risk or observation everyone else talked straight past.
Give AI context instead of starting from scratch
This was probably the biggest improvement I made.
I keep a running project context with the basics: objective, scope, current phase, milestones, stakeholders, decisions already made, current risks and issues, dependencies, constraints and priorities.
Then I use the relevant parts of that alongside my prompts rather than explaining the project from zero every time.
The difference in the output is huge.
It also makes AI useful for much more than writing. I can give it last week’s project position and this week’s information and ask:
“What changed? What was missed? What became a risk? What was resolved? Is anything contradicting what we previously reported? What needs my attention?”
Sometimes my prompt is simply:
“Don’t summarize everything. Tell me what I need to know as the PM.”
That’s where AI starts becoming genuinely useful. Project managers deal with huge amounts of information. Sometimes the difficult part isn’t writing the status report; it’s spotting the three things hidden among 50 updates that actually matter.
Obviously, be sensible about what information you provide. I only use information I’m permitted to use with tools approved by my organization. Confidential client information, personal data and commercially sensitive information shouldn’t simply be pasted into a public AI tool.
Use it to write less – and think more
Once a meeting is finished, I’ll use AI to turn the transcript and my own notes into a recap: decisions, actions, owners, dates, dependencies and anything still unresolved.
I can then take the same verified information and adapt it for different audiences. The working team might need the detail. A sponsor wants progress, risks and decisions required. A director might only need the outcome, impact and biggest blocker.
I’m not asking AI to create different versions of reality. I’m asking it to communicate the same facts differently.
But increasingly, some of my favourite prompts don’t ask AI to write anything at all.
Before a difficult meeting I might ask:
“What questions is the sponsor likely to ask me?”
“What am I assuming that might not actually be true?”
“If you were the technical lead, what would concern you about this plan?”
“What risks or dependencies might I be overlooking?”
I don’t automatically agree with the answers. That’s not the point. I’m using it to pressure-test my own thinking.
I do something similar with risks. AI can produce a useful first-pass list of risks based on the type of project, but nothing goes into my RAID log simply because AI suggested it. I still decide whether it’s relevant, how serious it is and what we should actually do about it.
My rule is pretty simple: AI drafts. I decide.
AI still needs a PM
AI transcription and summaries are good, but they’re not infallible. Names get muddled, numbers get misheard and nuance disappears.
I still keep very short handwritten notes during important meetings — decisions, dates, risks and anything I know I’ll need to check afterwards.
But I don’t think responsible AI use means spending half an hour checking every sentence it produces either.
I pay particular attention to anything that can materially change the project if it’s wrong: dates, numbers, owners, decisions, dependencies, commitments and anything being escalated.
If someone says, “I’ll try to get that done Friday”, and AI turns it into “John will deliver this Friday”, that’s not a wording difference. It’s changed the commitment.
Those are the things worth checking carefully.
And this becomes much quicker with practice. You learn what your AI tool tends to get wrong, your prompts improve and you start reusing instructions that work.
For example:
“Separate confirmed decisions from things that were only discussed.”
“Don’t assign an owner unless one was explicitly agreed.”
“Flag anything you’re uncertain about rather than filling in the gap.”
That’s when the time saving really starts to show.
AI doesn’t replace the project management basics
AI can’t magically fix a project where nobody knows the actions, risks, owners or next steps.
I still want a clear lifecycle. I still want my RAID log. I still want actions with owners and responsibilities that are clear.
AI sits around that structure and makes some of the work faster. It doesn’t replace it.
That’s also why I created my PM Smarter Project Lifecycle Checklist & Tracker. It keeps the lifecycle checklist, dashboard, RACI, RAID and actions together so there’s a proper project structure behind all of this in the first place.
[View the PM Smarter Project Lifecycle Checklist & Tracker on Etsy → https://www.etsy.com/uk/listing/4550443008/pm-smarter-project-checklist-tracker ]
So, should PMs be worried about AI?
I think we should take it seriously, but I don’t think the most useful question is whether AI will replace project managers.
Project management isn’t just producing plans, meeting notes and status reports. It’s judgement, navigating ambiguity, getting people to make decisions, spotting when something doesn’t add up and knowing when something needs escalating.
AI can take a surprising amount of the work around that off our plates.
And I’d rather let it handle some of the repetitive admin so I have more time for the parts of the job that actually need me.
I don’t think PMs who embrace AI suddenly become replaceable. I think they become more efficient.
And I suspect the bigger career risk over the next few years isn’t losing your job directly to AI. It’s being compared with someone doing the same job who has learnt how to use it well.
So I’m learning, carefully. I’m keeping the judgement and checking what matters, but I’m absolutely using it.
Because once you’ve found the tools that work for you, refined a few prompts and built AI into the things you already do every week, it stops feeling like another technology you’re supposed to learn.
It just makes the job a little easier.
And right now, I’ll happily take that.