AI for meeting effectiveness is one of the most practical places for an EA to start using AI tools, because we already manage meetings from beginning to end, usually long before our Executive walks in. We pull the board pack together and chase missing papers the day before; we brief our Executive on what is coming up and what they need to decide; and afterwards we turn the meeting notes into minutes, actions, and follow-up. It is a huge amount of work, and most EAs do it really well without anyone seeing how much goes into it.
Most of the conversation about AI and meetings so far has focused on note-taking, and AI tools have got very good at it. But if you think about where your time in a meeting actually goes, you spend most of it preparing for it, documenting it, and following up afterwards, with the meeting itself only a small part of the whole process. That is where AI has moved the most over the last year, and where I think EAs have the most to gain.
This isn’t just another AI list; it’s a collection of prompts written by Assistants, for Assistants, to help you get the most out of ChatGPT and start making a real difference in your day-to-day.
It helps to look at how quickly AI has progressed, because I think many EAs still use AI for meetings the way they started a couple of years ago, and the tools have changed a great deal since then. I tend to describe it in three stages.
Before AI, everything around a meeting was done by hand. You sat in the meeting writing as fast as you could while following the discussion (and keeping an eye on whether your Executive needed anything), then typed the notes up afterwards, worked out who had agreed to do what, formatted the minutes and emailed them round, hoping people would actually read them. Preparation worked the same way. You read the board pack yourself, highlighted what mattered and wrote the briefing from scratch. It worked, and plenty of EAs became extremely good at it, but it took hours, and the quality of the meeting record depended on how much you could write down while doing everything else.
The first wave of AI arrived in two parts. Transcription tools, such as Teams and Google Meet transcription, Otter, Fathom, and Granola, started listening to meetings and producing a transcript, a summary, and a list of action points so we could focus on the conversation and the people in the room instead of scribbling. Around the same time, chat tools like ChatGPT, Claude, Copilot and Gemini meant we could paste in a transcript or upload a document and ask for something back. Summarize this. Pull out the actions. Draft the minutes. Rewrite this follow-up so it sounds more like my Executive.
That was a big step forward, and it is where most of the EAs I speak to are now. The limitation is that it works one request at a time. You ask, it answers, you copy the answer into another document, and then you ask the next thing. You are still the person moving information between the transcript, the action tracker, the inbox and next week’s agenda, and the tool only knows what you have pasted into that particular conversation.
The third stage is agentic AI. You’ll hear “agentic” a lot over the next year; it simply means the AI can work toward an outcome you give it, decide the steps itself, and use the apps and documents you have given it access to, instead of waiting for you to prompt each step.
For meetings, that makes a real difference to how the work gets done. The tool can work across several sources at once, so it can read the board pack, last month’s minutes, your action tracker and the emails sent since the last meeting in one go, as long as you have connected them. You can give it the whole assignment, such as “prepare my Executive for Thursday’s board meeting”, and it works through the steps and returns the finished outputs for you to review. It can also update documents you already have and run on a schedule, so the action tracker updates after the meeting and gets checked again on Friday without you starting a new conversation each time. Most of these tools will ask before sending anything or changing a file, and you decide which apps and documents they can see, which matters a lot when you are working with your Executive’s information.
These tools are not perfect, and everything they produce still needs checking, but they can now take on a piece of work with several steps in it, which is exactly what meeting preparation and follow-up usually is. That is what makes AI for meeting effectiveness worth the time. You can hand over a piece of work in the same way you would brief a colleague: here is the job, here are the documents, here is what I need at the end, and here is what you must not do.
A board meeting is a good place to start, because if you have ever supported one, you know the work begins weeks before anyone sits down. The agenda, board pack, and individual papers include financial information, recommendations, and risks spread across them, and our job is to understand enough to ensure our Executive is properly prepared.
So the first step in using AI for meeting effectiveness is understanding what is in the pack, before you ask it to produce anything. Upload the papers, or connect it to the drive where they live, and ask it questions. Which papers need a decision, and which are for discussion or information only? Are there any significant financial commitments or risks? Do any two papers give different figures for the same thing? Which three papers should your Executive spend the most time on before the meeting, and why? At this stage, you are exploring the material and deciding where to look, and you can keep going until you understand what you are dealing with.
Once you understand the pack, you can switch from asking questions to handing over the assignment. Here is a version of the prompt I use in the masterclass:
“Prepare my Executive for the upcoming board meeting, using the board pack and individual papers in our shared drive. Create a one-page briefing on the five issues they most need to understand, a summary of every decision the board is being asked to make, a list of questions they may want to ask in the meeting, and a checklist for me of anything I need to confirm, chase or prepare beforehand. Use the papers as the source of truth. Do not invent missing information or treat a discussion item as a confirmed decision, and flag anything unclear or inconsistent for my review. Do not send or share anything.”
The boundaries at the end of that prompt are the most important part of it. AI tools are very good at filling gaps with something plausible, so you need to tell them to show you the gaps instead. If a paper recommends spending $50,000 but doesn’t clearly explain the return, or a decision has no deadline attached, you want that flagged before your Executive gets into the room, not discovered halfway through the meeting.
Then you review everything, and this is where your knowledge of your Executive comes in. You check the figures and the recommendations, and you check that the tool has understood the papers correctly. You might also find that the issue it put fourth on the briefing is the one you know your Executive will see as most important on the agenda, so you move it to the top. We’ve written more about checking information before it reaches your Executive, and the same habits apply to anything AI has produced for you.
The checklist it creates is often the most useful output of the four, and a good example of AI for meeting effectiveness that helps you and your Executive. A missing paper, a figure somebody still owes you, a decision that depends on information that hasn’t arrived yet; these are the things you want to know about on Monday, not on the morning of the meeting.
During the meeting, AI for meeting effectiveness mostly gives you your attention back. When you are not trying to write everything down, you can watch the room, notice who has gone silent on a topic, and catch the comment that never makes it into the formal minutes.
There are plenty of tools that will capture the conversation for you, and most organizations will already have one or two of them available:
Some of these tools also let you ask questions while the meeting is still going on, without interrupting anyone. You can check what was agreed on the venue budget twenty minutes ago, or get a quick summary to catch up someone who joins late, while the discussion continues around you.
Make sure everyone in the meeting knows AI is being used before you switch anything on, and check what your organization allows, because some environments (especially heavily regulated ones) limit which tools can be used and where the recordings are stored.
After the meeting is where most EAs lose the most time, and where AI for meeting effectiveness pays back fastest.
Most of us, when we first start using AI here, prompt for each thing separately. Summarize the transcript. Now pull out the actions. Now write the minutes. Now draft an email to the attendees. That works, but it is five conversations for one job. Instead, give the tool the transcript alongside your agenda, your meeting template and your existing action tracker, and ask it to complete the whole process in one go: the minutes in your usual format, the decisions that were actually confirmed, and the actions grouped by owner.
Two things make a big difference to the quality of what comes back.
Ask it to update your existing documents instead of creating new ones.
A fresh action list after every meeting is how actions get lost. You want one master tracker, so the tool can see what was already outstanding, mark what has now been completed, and add what is new. The same goes for decisions, because a decision log that builds up meeting by meeting saves a lot of digging when someone asks, three months later, “When did we decide that?”
Ask it to find anything you were given.
We all know how this happens. Your Executive says, “Ask [your name] to get some dates,” or “She’ll have that information,” and you were never in the room to hear it. Thank you very much. A separate list of everything delegated to you, pulled out of the transcript, means you are not watching a 45-minute recording looking for your own name.
Then you review it all, because AI is processing language here and it can get it wrong. If the minutes say your Executive approved something, make sure they approved it and didn’t just say it sounded sensible. If someone has been given a deadline, check it was agreed rather than mentioned in passing. It is a very strong first draft, but it is still a first draft, and we’ve written about why filtering AI output is becoming a core EA skill.
Board meetings might happen only a few times a year, but most of the meetings we manage happen every week, fortnight, or month. Your Executive’s 1:1s with their direct reports, the leadership meeting, the project meetings, the regular finance review. Every time one happens, new actions and decisions emerge, and items from last time either get resolved or need to come back.
This is where AI for meeting effectiveness starts to save you real time, because you can build the process once and reuse it.
The first step is to give each recurring meeting its own home. In ChatGPT and Claude, that is a Project. Copilot has Notebooks that work similarly and can now pull in Teams meeting transcripts and notes directly; in Gemini, the closest equivalent is a Gem, which holds instructions and files you can return to. Add instructions about who attends and how the meeting should be managed, and upload the permanent documents: your meeting template, the master action tracker and the decision log. Add your boundaries here too, so they apply every time; don’t invent an owner or a deadline, don’t turn something that was discussed into something that was agreed, and flag anything unclear. As meetings happen, that workspace builds the relationship history, which matters because this week’s meeting is almost always connected to what happened a fortnight ago.
The second step is to make your after-meeting process reusable. ChatGPT and Claude both call these Skills, and you don’t need any technical knowledge to create one. You describe your process in plain language, the way you would explain it to someone covering for you, and the tool turns it into a set of instructions it can follow every time. When I build these in the masterclass, the tool usually comes back with questions. How detailed should the minutes be? What should happen when an action has no deadline? How should completed actions be recorded? Those are good questions to be asked, because they make you write down a process that has probably only ever existed in your head.
Once it is set up, processing the meeting can be one line: “Process the latest 1:1 between my Executive and the General Manager.” You still review everything before it goes anywhere, but you are no longer working through every administrative step by hand. This is the part of using AI for meeting effectiveness that I would encourage you to talk about with your Executive, because you have taken a process that relied entirely on you and turned it into a documented system that works the same way every time, and that is a good example of the value you bring to the role.
The weeks between meetings are where actions go to die (and every EA has chased at least one of them back to life). So the last stage of AI for meeting effectiveness is making sure everyone actually does what they agreed to.
ChatGPT and Claude both let you set up scheduled tasks, which run on their own at a time you choose. Copilot calls them scheduled prompts, and the Gemini app has scheduled actions, so most of you will have some version of this available. For a recurring meeting, a weekly check works well. Every Friday morning, for example, the tool reviews the master action tracker and tells you what is overdue, what is due in the next seven days, what looks blocked and which decisions are still waiting for information. It can draft the follow-up messages for you, but set it up not to send them, because the decision about who to chase, and how firmly, is still yours.
You can take this one step further before the next meeting. Your Executive and their direct report don’t just talk in their 1:1; they email each other in between, questions get answered, new issues come up, and sometimes they make a decision by email that never makes it into the tracker. If your email is connected, you can ask the tool to review the emails between them since the last meeting, compare them with the meeting record and the action tracker, and draft the next agenda, with the source of each item listed so you can check it. Anything resolved by email drops off, so meeting time goes to what still needs discussion.
Over a few months, using AI this way also helps spot problems before your Executive does. When the same action has slipped three meetings in a row, or the same topic keeps coming back to the agenda without being decided, you can spot the pattern and raise it while it is still easy to fix.
You don’t need to set up AI for meeting effectiveness across every meeting at once, and I would really encourage you not to try.
Pick one recurring meeting, ideally one you support every week or fortnight, so you get plenty of practice quickly.
Write down everything you do after that meeting, including the checks you make and the things you look out for that nobody has ever asked you to.
Turn that list into a single assignment, run it three or four times, and adjust it each time it gets something wrong.
Once it is producing what you need, make it reusable with a Project and a Skill, and add a weekly check on the actions.
Then move on to the next meeting, and the one after that. AI for meeting effectiveness builds up one meeting at a time, and each one you set up is a meeting your Executive walks into knowing exactly what they need to decide.
Share this article:
Join 30,000+ Assistants receiving our carefully crafted weekly newsletter packed with valuable tips, tricks, and insights tailored specifically for Assistants like you.
