The question started small: could two systems talk to each other?
Once the Notion project management system was running, a new kind of problem became visible. We had better project tracking, but some of the data still had to be collected manually from another system. Every week, PMs would go through projects in the meter data portal and update Notion so the commissioning progress stayed current.
Across the team, that took hours. It was not complex work, but it was repetitive and easy to delay because it required opening another system, checking each project, interpreting the meter information, and then entering the update somewhere else.
One day I had the thought: what if I could pull the data from the portal and push it into Notion automatically?
I asked ChatGPT a question and fell into the rabbit hole.
I started asking ChatGPT whether the idea was even possible. That led me to APIs, API keys, documentation, endpoints, response data, and all the vocabulary that felt completely foreign at first. I requested access and documentation from the company that hosted our meter data, then started reading what points were exposed.
The first goal was not to build the whole automation. It was just to see if I could pull anything. I asked for help writing a simple Python script, added the API key, called one endpoint, and waited to see what came back.
It worked.
That sounds small now, but at the time it felt like magic. Data from a system I used every week was suddenly available in my terminal. I remember feeling shocked and excited because my brain immediately went to the next thought: if this is possible, a lot more is possible.
The test became a scheduled automation.
After the first pull worked, I started shaping the data into something useful. The script needed to know which meter points were commissioned and which were not. It also needed to sort the counts by commodity, so I built rules for water, thermal, hydro, EV, and gas meters. From there, I connected the Notion API and tested whether those summarized counts could update the right project records.
That worked too. Eventually the script became a weekly automation that runs every Monday at 10:30 AM. It processes 189 projects, pulls the meter data, applies the commissioning logic, and pushes the project-level progress into Notion. The update only moves one way: from the meter data portal into Notion.
Before this, four PMs spent roughly eight hours a week keeping those progress bars current by hand. The automation removed that manual update cycle, but the bigger change was personal. This was the project where I learned Python, APIs, request sequencing, data shaping, and how to avoid hammering a server. It was also where I realized AI tools could help me build things I did not yet know how to build.
- Runs every Monday at 10:30 AM.
- Processes 189 projects per run.
- Saves roughly eight hours a week across four PMs.