Automating My Rental Property Bills with a Personal AI Assistant

ai automation openclaw personal-productivity

Start With Your Own Problem

There’s a principle I keep coming back to when thinking about AI projects:

The best AI project is one that solves your own problem first.

Not a hypothetical user’s problem. Not a market opportunity. Your problem — something that currently annoys you, wastes your time, or requires repetitive effort you’d rather not do.

Why? Because you deeply understand the problem. You know the edge cases. You’ll actually use it. And most importantly, if it’s valuable to you, there’s a good chance it’s valuable to others like you.

This is the story of one such project.

The Problem: Rental Property Admin is Tedious

I own a rental unit at Astoria Ampang. Every month, two utility bills arrive:

A TNB electric bill (emailed from billing.oms@tnb.com.my)
A Water bill from CSS (emailed from no-reply@csshome.info)

My workflow used to look like this:

Log into Gmail, search for the TNB email
Download the PDF attachment
Rename it properly
Move it to the right Dropbox folder (Astoria B1002 Bills/YYYY-MM/)
Repeat for the water bill
Open both PDFs, find the amounts
Manually calculate the payout template and send it to my rental operator

On the 20th of every month, I also send this message to my rental operator to request the revenue payout:

RM2700 x 80% = 2160.00
Water RM56.70
Electric RM248.40
Internet 102.80

Total Revenue for October 2025
RM2567.90

This took maybe 15 minutes each month. Not much on its own — but it’s exactly the kind of low-value, repetitive task that adds up and drains mental energy.

The Solution: Let the AI Handle It

I set up OpenClaw as my personal AI assistant running locally on my Mac, connected via WhatsApp. Then I automated two things:

1. Astoria Bills Sync

When I say “sync Astoria bills” (or it runs automatically on the 5th of each month), the assistant:

Checks which month folders in ~/Dropbox/Astoria B1002 Bills/ are missing bills
Searches Gmail for TNB electric bills (from:billing.oms@tnb.com.my)
Searches Gmail for CSS water bills (from:no-reply@csshome.info)
Downloads the PDF attachments directly into the correct Dropbox folder
Reports back what was downloaded

The key tool here is gog — a Google CLI that lets you interact with Gmail programmatically. The AI uses it to search threads, list attachments, and download them — all from the command line.

2. Astoria Revenue Calculator

On the 20th of every month at 9am, the assistant automatically:

Determines which month to bill for (always 3 months prior — so January 20 → October bills)
Finds the TNB and water PDFs in Dropbox for that month
Extracts the bill amounts using pdftotext
Calculates the total: (RM2700 × 80%) + water + electric + RM102.80 internet
Formats and sends me the ready-to-forward payout message via WhatsApp

The extraction uses a simple Python script that parses the PDF text and pulls out the specific amounts. Here’s the core logic:

def extract_tnb_amount(folder):
    for f in os.listdir(folder):
        if f.startswith("OPC2_") and f.endswith(".pdf"):
            text = pdf_to_text(os.path.join(folder, f))
            idx = text.find("Jumlah Bil Anda (RM)")
            if idx != -1:
                amounts = re.findall(r'(\d{1,4}\.\d{2})', text[idx:idx+200])
                if amounts:
                    return float(amounts[0]), f

Not glamorous — but it works reliably every month.

Why This Works as an AI Project

A few things made this a good fit for AI automation:

1. The data is structured but scattered Gmail, PDFs, Dropbox — it’s all there, just not connected. An AI assistant with the right tools can bridge these without building a custom integration.

2. The logic is simple but tedious The calculation is straightforward. The hard part is the repetitive searching, downloading, and formatting — exactly what automation is good at.

3. The value is immediate and personal I use this every month. I notice immediately if something breaks. That feedback loop makes the system better over time.

The Broader Lesson

I could have spent weeks building a polished web app to manage rental properties. Instead, I spent a few hours teaching my AI assistant a workflow I already do manually.

The difference:

Traditional approach: Build for an imagined user, then find out if it’s useful
Personal AI approach: Solve your own problem first, then decide if it’s worth sharing

The second approach ships faster, fails cheaper, and almost always produces something more honest and useful.

Start with your own problem. Make it work for you. Everything else follows.

Tech Stack

OpenClaw — personal AI assistant running locally on macOS
Claude Sonnet — the underlying model
gog CLI — Google CLI for Gmail operations
pdftotext — PDF text extraction
Python — glue for the revenue calculation script
Dropbox — cloud storage for the bills
WhatsApp — the interface I actually use daily

The entire setup lives in ~/.openclaw/workspace/ — memory files, scripts, and all.