My executive assistant used to work 20 hours a week for me.
She still works for me. She’s still great at her job. But it’s 3 to 4 hours now.
What happened to the other 16 hours? I built AI agents to handle them. Specifically, four workflows that turned out to account for almost all of her time.
This isn’t a story about replacing an EA. It’s a story about figuring out which parts of assistant work actually need a human — and which parts are coordination and admin that AI handles better anyway.
The Four Workflows
When I looked at what my EA was actually doing across 20 hours a week, it clustered into four areas:
1. Email inbox management
The single biggest time sink. Sorting through what came in. Figuring out what needed a response. Drafting replies. Flagging things that needed my attention. Archiving what didn’t.
I now have an agent that handles most of this. It sorts incoming mail, drafts replies in my voice, surfaces what genuinely needs a decision, and archives the rest. She reviews what the agent surfaces — which takes a fraction of the time that processing from scratch did.
The math is simple: five to ten minutes saved per email, times 50+ emails a day, adds up to hours every week.
2. Meeting prep
Before every important meeting, someone had to do the research. Pull up the email thread. Look up who the person was. Check what was outstanding. Compile it into something usable.
That’s now done by an agent that runs before every meeting on my calendar. It goes through my email history with the person, does background research, reads their LinkedIn, and generates a brief. My EA reviews it. I read it. The meeting prep that used to take 20-30 minutes per call is now done in under two minutes.
3. Meeting notes and follow-ups
After every call, something had to happen with the recording. Notes had to be created. Follow-up emails had to be drafted. Action items had to be captured somewhere.
An agent now handles the first pass on all of this. It transcribes, generates structured notes, drafts the follow-up email, and creates task items. My EA reviews the output and adjusts what needs adjusting. The whole process that used to take 30-45 minutes per meeting now takes a few minutes.
4. Scheduling
Calendar back-and-forth is one of those things that feels simple but accumulates time fast. “Are you free Thursday?” “I have a conflict Thursday, how about Friday?” This kind of coordination — happening across ten or fifteen threads simultaneously — is exactly what AI handles well.
Most scheduling is now automated. My EA handles the edge cases: meetings where I have context she needs to know, relationships where a direct human touch matters, situations where the calendar logic gets complicated.
What’s Left for the Human
The 3-4 hours that remain aren’t the leftover scraps. They’re the things that genuinely require a person.
Judgment calls where she knows my priorities and relationships well enough to decide on my behalf. Situations where the email needs a specific tone she knows how to calibrate. Scheduling scenarios where someone important to me is involved and she knows that context. Anything where someone might feel the difference between AI-handled and human-handled.
That’s where a great EA creates value. Not in processing email. Not in calendar logistics. In judgment and relationship-sensitive situations where the human touch actually matters.
The mistake I made for years was paying for both. I had someone spending most of her time on repetitive coordination work — which, if I’m honest, she didn’t love — and only a fraction of her time on the higher-judgment work where she’s genuinely excellent.
The Map-Before-You-Decide Framework
If you have an EA, or you’re thinking about getting one, the question worth asking first is: what does the role actually consist of?
Write down every task that role touches in a week. Then ask, for each one: is this repetitive and rule-based, or does it require genuine human judgment?
The repetitive, rule-based work — processing incoming communications, scheduling, generating first drafts of notes and follow-ups, doing research and prep — is AI’s job. AI is available 24 hours a day, doesn’t need context on every edge case explained, and handles the volume without fatigue.
The judgment-heavy work — navigating relationships, making decisions on your behalf, knowing when something is sensitive — is where you want a human.
Most people who work with executive assistants are inadvertently paying for both in the same package. They don’t have to be.
What This Looks Like in Practice
If you’re starting from zero, the fastest path is picking one of the four workflows and automating it completely before touching the others.
Start with email if you’re buried in your inbox. Start with meeting prep if you consistently walk into calls underprepared. Start with scheduling if the back-and-forth is driving you crazy.
Each one has a clear before and after, and the setup is simpler than most people expect. Once the first one is running, you’ll have both the experience and the motivation to build the next.
My EA still works with me. She does things AI can’t do. But what she spends her time on has fundamentally changed — and honestly, she’s told me she prefers it. Nobody enjoys processing email for four hours a day. Getting to spend her hours on work that actually requires her judgment? That’s a better job.
20 hours to 3-4. The work didn’t go away. It just moved to the right place.
Recommended for you
Want the full system? 25X is the flagship productivity system we teach.
Explore 25X →