Welcome to everyone joining us for the first time, and welcome back to our regular readers!
We missed you last week! We used the extra week as an opportunity to spend some time with your survey feedback and think more intentionally about what you want from this newsletter. One thing came through clearly: you want practical ways to take AI beyond the basics and put it to work in your day-to-day role as an EA.
If you’re new here, that’s what this newsletter is about: practical, useful ways to make AI part of your everyday work as an EA, not just another tool to experiment with.
So this week, we’re digging a little deeper.
We have more ways than ever to capture what happens in a meeting. Teams, Zoom, AI note takers, transcripts, and our own notes leave us with plenty of documentation. But having meeting notes is not the same as having a follow-through system. Notes tell you what was said. A follow-through system tells you what still needs to happen. For an EA, the most useful information is often buried inside all those notes:
What was actually decided?
Who committed to doing what?
When is it due?
What question never got answered?
Who needs a follow-up?
And perhaps most importantly, what did everyone agree to last week that somehow disappeared from this week’s conversation?
This week, we’re skipping the meeting summary. Instead, we’ll use AI to turn messy notes into something you can actually manage. The workflow is simple:
Meeting notes → Extract → Verify → Track → Follow up.
Step 1: Start with the notes you already have
You do not need to change your meeting platform or start using a new AI note taker for this workflow. Start with whatever you already have after a meeting. It could be a transcript, AI-generated meeting notes, your own typed or handwritten notes, or a combination. Your AI tool might accept a photo of your written notes! Before uploading anything, make sure the content is appropriate for the AI tool you're using and follows your organization's policies for confidential and sensitive information. When in doubt, ask before you upload.
Then, give AI a more specific job. Instead of asking “summarize this meeting”, ask it to “tell me what happens next.”
Step 2: Extract the things we actually need
This is where we introduce an AI skill called structured extraction. It sounds technical, but the idea is simple. You're taking unstructured information, such as several pages of meeting notes, and asking AI to pull out specific information and organize it into a structure you define. The more clearly you define that structure, the more useful the output becomes. For this workflow, we want decisions, commitments, owners, deadlines, unanswered questions, and follow-ups.
Prompt: Turn meeting notes into follow-through
Copy and customize this prompt:
You are helping me manage follow-through after a meeting. Review the meeting notes below and extract:
Decisions that were made
Action items or commitments
The owner of each action or commitment
Any deadline or timing that was explicitly mentioned
Unanswered questions
Follow-ups that are needed
Items that require executive attention
Organize the results into a table with these columns:
Type | Item | Owner | Deadline | Status | Follow-Up Needed | Source/Evidence
Important: Do not guess or infer an owner, deadline, decision, or commitment that is not clearly supported by the notes.
If information is missing, write “Not specified.”
If something needs clarification, flag it rather than filling in the missing information yourself.
Keep the Source/Evidence field brief so I can quickly verify each item against the original notes.
MEETING NOTES:
[PASTE YOUR NOTES HERE]Step 3: Verify before you act
Do not copy AI’s output straight into your task manager and start sending reminders. Take a few minutes to compare the output with your original notes or transcript, especially these four fields: Decisions, Commitments, Owners, and Deadlines. A wrong owner or deadline can send a reminder to the wrong person at the wrong time.
AI-generated meeting content can be incomplete or inaccurate. Microsoft, for example, advises users to verify AI-generated meeting content rather than assuming the output is correct. Think of AI as the first pass. You are still quality control. The Source/ Evidence column in our prompt makes this easier because you can quickly see what part of the notes led AI to create each item. Once you've verified the output, it becomes operational information. Now you can act on it with confidence.
Step 4: Turn the output into your EA follow-through tracker
Your meeting transcript might be ten pages long. Your working tracker shouldn't be. Once the key details are verified, simplify them into a tracker like this:
Commitment | Owner | Due | Status | EA Next Step |
Send revised proposal | James | Friday | Open | Check Thursday afternoon |
Review Q4 budget | Aiden | Not specified | Open | Confirm desired deadline |
Schedule client follow-up | Lola | Next week | Open | Send scheduling options |
Notice the last column. EA Next Step is where your judgement comes in. It turns each item into a specific action you can take. This is the shift we're making: Stop managing meeting notes. Start managing commitments.
Your tracker can live wherever you already work: a spreadsheet, project management system, CRM, or whatever tool your team already uses. The best tool is the one you will actually check.
Step 5: Let AI help you draft the nudges
Now we reach a familiar EA task.
The deadline is approaching. Something still isn't done. You need to follow up without sounding robotic, impatient, or like you're policing everyone else's work. Rather than drafting every reminder from scratch, give the verified items from your tracker back to AI.
Prompt: Draft the follow-ups
Using the verified follow-through items below, draft a brief follow-up message for each item that needs action.
Keep each message friendly, professional, and natural.
Include the relevant commitment and deadline only when they are provided in the tracker.
Do not invent new context, deadlines, commitments, or urgency.
Make each message sound like a helpful EA follow-up, not an automated reminder.
VERIFIED ITEMS:
[PASTE ITEMS HERE]The prompt works even better when you add context about the recipient and channel. For example:
This message is going to my executive in Teams. Keep it conversational and brief.
Or:
This is an external email to a business partner. Keep it warm and professional.
AI gets you the first draft faster. You still decide what gets sent. You know the relationships, the timing, and the tone that works for each person.
Level It Up: Find What Quietly Disappeared
Here's where this workflow gets more interesting. Most executive work doesn't happen in a single meeting. Projects stretch across weeks. Decisions evolve. Deadlines move. New priorities arrive. Something that received ten minutes of discussion last week may not get mentioned at all this week.
But “not mentioned” doesn't mean “completed”. Items that drop out of the conversation can be the easiest ones to lose track of.
So after a recurring meeting, don’t start from scratch. Keep the verified tracker from the previous meeting. After the next meeting, give AI two things: your previous verified tracker and your new meeting notes. Then ask it to reconcile them. Now AI isn’t just processing one meeting. It’s helping you spot gaps across meetings.
Prompt: Compare this meeting with the last one
Compare the previous meeting's verified follow-through tracker with this week's meeting notes.
Review each previous commitment and categorize it as:
Completed: The new notes clearly indicate that it was resolved or completed.
Carried Forward: The item is still active.
Changed: The owner, deadline, scope, or decision has changed.
Missing: The item was previously open but is not addressed in the new meeting notes.
Unclear: There is not enough evidence to determine its current status.
Then identify any New Commitments created during this meeting.
Do not assume an item was completed simply because it was not mentioned again.
Do not invent status updates or infer completion without evidence.
Present the results in a table with:
Item | Previous Status | Current Status | Owner | Deadline | What Changed | Follow-Up Needed | Source/Evidence
Flag anything that requires human clarification.
PREVIOUS VERIFIED TRACKER:
[PASTE HERE]
CURRENT MEETING NOTES:
[PASTE HERE]Pay close attention to anything marked Missing or Unclear. Those items are your next follow-up conversations.
The AI Skill Behind This: Structured Extraction
If you followed this workflow, you practiced an AI skill called structured extraction. Instead of asking AI an open-ended question and receiving another wall of text, you defined what information you wanted and exactly how you wanted it organized. That's useful far beyond meeting notes.
You can apply the same approach to an email thread, project update, event planning document, executive briefing, client notes, or any other pile of information that needs to become something actionable.
The formula is straightforward:
Give AI the source → Define what to extract → Define the output structure → Tell it not to guess → Verify the result.
And once that information is structured, you can do more with it. You can track it, compare it, find gaps, draft communications from it, and carry unresolved items forward.
Try it this week with one recurring meeting. Then run the comparison prompt after the next one.
AI Updates of the Week
Google is rolling out expanded Gemini capabilities across Gmail, Drive, Docs, Slides, and Google Chat. Rather than requiring you to use Gemini inside a single app, you can ask it to create a document or tracker, research material across files and threads, draft an email, schedule a meeting, or turn a request into Google Tasks without changing applications. The rollout began September 2 and may take up to 15 days to appear.
Google also announced the Gemini desktop app for Windows 10 and 11 in its September 11 Workspace update recap. Google describes it as a way to get help alongside the tools you already use. Press Alt + Space to bring Gemini up over whatever you are working on. For Workspace accounts, it’s on by default wherever Gemini is enabled.
Microsoft is expanding AI model choice inside Copilot. Grok models are now entering a limited Frontier preview in Word, Excel, and PowerPoint. Organizations must enable access through an admin setting first - that setting is off by default. For EAs, the lesson is simple: more AI options may be coming inside Microsoft 365, but privacy and approval rules still come first.
We’d still love to hear from you!
If you haven’t taken our 2-minute survey yet, there’s still time. Tell us how you’re using AI, where it’s falling short, and what you want to learn next. Your responses help us understand what EAs actually need from AI.
See you next week!
And please keep sharing what you’re learning with us. The more we hear from EAs doing this work every day, the more useful these conversations become for everyone.
The Base Team
