I built an AI executive assistant over a weekend. It changed how I work.
I built an AI executive assistant over Easter weekend, and it has already transformed how I work.
I'd been hearing about people building AI chiefs of staff for several weeks. But it always seemed way beyond what I could do.
Then I found a tutorial by Chris Blattman with step-by-step instructions for getting started with Claude Code. It was clear enough that I could follow along without getting stuck. His approach is top down: define tasks, write detailed prompts, and give Claude clear structure. That was exactly what I needed. A clear starting point.
Then I listened to the podcast "How I AI" with Claire Vo, where she interviewed Hilary Gridley, who has a similar system built with a bottom-up approach. She started with fewer directions, uses Claude Code naturally, and lets the system learn her preferences over time.
I decided to build something in the middle. I created context files about me, my business, my clients, my preferences, and my personal life. These were all built with Claude's help, no code. It probably took me about six hours in total.
Here is what it does now:
- Plans my day. Every morning I say "plan my day." By the time I make my coffee, Claude has scanned my email, categorized everything, and made a recommendation for each one: file it, review it, draft a response, or leave it for me. I approve or adjust, and each time, it updates its own memory.
- Captures my voice notes. Throughout the day, whenever something comes to mind, I talk into a widget on my phone and the task gets captured. Every morning those tasks are imported automatically.
- Reads my meeting transcripts. It goes through the transcripts from the day before and pulls out every action item, whether it is mine or something I need to follow up on with someone else.
- Builds my schedule. It checks my calendar, asks if I need briefings before any calls, and runs the research if so. It knows I want 20% of my time on business development. It blocks 60 minutes on Tuesdays and Thursdays for my run. It factors all of this in and maps out my day.
Throughout the day I talk to Claude using Wispr Flow. I hit two keys and tell it what I'm working on, much like I would to a colleague sitting next to me. It updates tasks and calendar as I go, and flags when something is overdue.
On Fridays, I run a weekly review. It summarizes how I did against what I set out to do: whether I gave my clients the focus they needed, whether I dedicated enough time to my own business development, and observations about how the system itself could be improved.
The things that used to live in my head, or sit in a Trello board I rarely opened, are now raised each morning. The more it learns about how I work, the more useful it gets. This has become the backbone of how I run my practice.
This is far less complex than it sounds. You do not need to be technical. You just need to start, and when you get stuck, ask Claude to walk you through it step by step.
This is one piece of my AI operating system.
I write about building it, in public, in my newsletter AI on Purpose.
Subscribe to AI on Purpose