Reading time: 3 min
Watching time: 10 min (2x speed)
This week is a double build edition.
2x AI projects, solving the same annoying thing: the stuff that eats minutes of my day for no reason. I wanted to share them while they're still at 80%, as I'm continuously refining them.
What other things would you want automated?
Real ones that would really change your game.
We start with the problem in mind
In literature, there is “the crisis” moment that changes the course of a story.
For the projects today, these moments are:
→ Invoices landing in my inbox in different formats, and needing a redundant process to extract the data, save it to a spreadsheet, save the invoice to a folder, and archive the email.
→ A cluttered inbox that isn't really about reading, it's about sorting. Deciding what needs me, what's worth a look, what's noise and what emails actually hide cool opportunities I might miss.
I'd been doing the work manually, and I needed some help (ideally free).
So I built two separate automations using AI.
AI to (actually) use #1:
Invoice Processing Automation
Every month, invoices arrive in my inbox, and each sender has a different way of sending them:
→ One buries the amount inside a PDF.
→ One sends the same email address for two invoice types.
→ One doesn't attach the PDF at all, and it makes you log into a portal.
I mapped the patterns and used Google Apps Script to save invoices to Drive with clean names, log everything to a spreadsheet, and archive the emails. All running automatically, daily. And on the 20th of each month, it emails me the full report.
What I used: Google Apps Script, Gmail + Drive + Sheets, Gemini API (free tier), JavaScript written with AI.
The key insight: there is no single method that works for all senders.
The product decision is matching each sender to the right strategy.
→ Regex where regex is enough (to keep it simple and cheap).
→ Gemini, where the data is buried in the PDF.
AI to (actually) use #2:
AI Inbox Operator
Different problem, same energy. I was spending 30+ minutes a day just sorting email (and I'm sure it's not just me).
So I thought AI might be able to help me with the sorting. The system labels emails, kills the noise, and sends me a 30-second digest.
The system is deliberately dumb-simple:
🔥 Action → reply, decide, or do something
📚 Read → worth reading later
🗑️ Delete → noise, out of sight
People are prioritized.
Newsletters get sorted into Tier 1 (always keep), Tier 2 (maybe), Tier 3 (promos, bye).
The agent sorts, then reports, and I do a quick check (a bit of micromanagement on my end, but I wanna review the work for a while, until I'm confident everything runs smoothly).
What I used: an AI as the brain (Claude Code or Gemini), Gemini 3 to write the Python script, Gmail API via personal OAuth, two markdown files (SKILL.md + rules.md) for the logic and the learning layer.
The common thread
Observation → sitting with the mess long enough to see the pattern and understand what really needs to be solved. The AI is good and fast, but it can only be useful after doing the slow part of paying attention.
A small self-leadership note
Why is it so hard to stay still, quiet?
Is it just me, or do we all zoom through life, like crazy ions between electrodes?
We automate, but not to save time to rest and connect. Instead, we work even more, get more shallow and distracted.
We invest time in building automations, but not in building the space to actually enjoy the time we save.
Though I'm a big fan of automations (I shared two this week), I'm also a big fan of building the space where we don't lose ourselves.
And I just invite you to sit with this thought for a second… or just sit still.
Wish you a quiet week (and the right automations that save you time for what matters).
Silvia

