About six months ago I did a time audit on my week. The results were depressing. I was spending somewhere between 90 minutes and 2 hours per day on email โ not reading important messages, just managing the noise. Sorting, archiving, writing the same polite replies over and over, unsubscribing from things I subscribed to once in 2019.
Today, that's down to about 20 minutes. My AI agent handles the rest.
I'm going to walk you through exactly how I set this up, what works, what doesn't, and the specific tools I use. No fluff โ this is the actual implementation, not a theoretical overview.
Before getting into setup, let me be clear about what's automated vs what still needs me:
What the agent handles:
What I still do myself:
The split is roughly: agent handles 70โ75% of the volume, I handle the stuff that reaily matters.
The core of my setup is OpenClaw running as my persistent AI agent, connected to Gmail via the Google API. OpenClaw is an open-source agent runtime โ it's the thing that keeps the AI running 24/7, gives it memory, and lets it take actions on my behalf.
Here's the high-level architecture:
You don't need all of these to start. You could begin with just the email reading and summarization, and add more capability over time.
OpenClaw installs via npm โ one command and it's on your machine. The initial setup takes about 20โ30 minutes if you follow the guide. You'll configure it with an AI model API key (I use Claude) and a messaging channel (I use Discord, but Telegram works too).
Full setup walkthrough: firstagentsetup.com/
This is easier than it sounds. You go to Google Cloud Console, create a project, enable the Gmail API, download credentials, and point OpenClaw at them. The whole process takes maybe 15 minutes once you know where to click.
The guide covers this step in detail with screenshots.
This is where you define what the agent does with your email. OpenClaw uses "skills" โ small text files that tell the agent how to behave for specific tasks. My email skill tells the agent to:
I have a scheduled task that runs at 8am every morning. The agent reads my inbox from the last 12 hours, writes a "morning brief" that covers the 3โ5 most important threads, and delivers it to a private Discord channel. This takes me about 5 minutes to read and means I go into my morning already knowing what matters.
The setup won't be perfect right away. I spent the first two weeks adjusting what the agent considers "important" and refining the draft-reply templates. By week three it was accurate enough that I trusted the summaries without double-checking everything.
Before the setup:
After the setup:
That's roughly 80 minutes per day back. Over a work week, that's almost 7 hours. That's a whole work day returned to me every single week, which is more than I expected when I started.
The biggest win wasn't time โ it was mental load. I stopped having the constant low-grade anxiety of "I should check my email." The agent checks it. I just read the summary.
A few things that didn't work well initially:
Letting the agent send emails without my approval โ I tried this for a week. It occured to me too late that tone and wording matters a lot in professional contexts. Now the agent drafts and I approve. Much better.
Giving it access to everything at once โ I connected my personal inbox, work inbox, and two side project inboxes simultaneously. It was overwhelming to tune. Better to start with one inbox and expand once it's working well.
Not defining what "important" means โ The agent needs examples. In the skill file, I spent time explicitly defining which senders are always important, which topics always need my attention, and which things can always be auto-archived. That specificity makes a huge difference.
Honestly, maybe. If you only get 20โ30 emails a day and they're all meaningful, this setup is probably more than you need. A simple Gmail filter setup might serve you fine.
But if you're getting 80, 100, 150+ emails a day and you feel like your inbox is a second job โ this setup is worth every minute of the initial configuration. The upfront time investment is maybe 2โ3 hours. The ongoing return is 1โ2 hours per day. That math works out pretty fast.
I wrote a complete walkthrough that covers everything above in detail โ including the skill file templates, the Gmail API connection process, and how to set up the morning brief. If you want to stop drowning in email, the guide is the fastest way to get there.
The guide covers OpenClaw setup, Gmail integration, and the exact skill templates I use for email automation. Step-by-step, no coding required.
Get The Guide โ $19 Get The Kit โ $39