Today we tackled the long-delayed task of setting up the draft-generation flow for my old email address. Instead of rebuilding from scratch, we duplicated the working webuildforchange.com email zap and customized it for the old inbox. After adapting the labels, IDs, and LLM prompts, the system started auto-triaging emails and saving drafts without intervention. It marked a real psychological turning point: the oldest backlog was finally moving.
Today was a game-changer! We finally crossed the automation chasm and set up the draft-generation flow for my old liveoutlaw.com email address. It’s been a long time coming, and let me tell you, the moment we saw those drafts appearing was nothing short of euphoric. Instead of starting from scratch, we decided to duplicate our successful webuildforchange.com email zap and customize it for the old inbox. This approach allowed us to focus on what really mattered: adapting labels, IDs, and LLM prompts to fit the unique needs of the old account.
The best part? As soon as we hit that “run” button, the system began auto-triaging emails and saving drafts without any intervention. It felt like a psychological turning point; the oldest backlog was finally moving! I mean, who doesn’t want to hear,
“OMG GUESS WHAT.... I'm actually getting through my emails! I ACTUALLY GOT THROUGH ALL MY DRAFTS!!!!”
That’s the kind of joy we live for!
You’d think duplicating a zap would be straightforward, right? Well, we hit a few bumps along the way. For starters, there was some confusion about whether Zapier could easily pull fresh label IDs again. Spoiler alert: it couldn’t. So, we found ourselves manually copy-pasting label IDs, which was tedious but necessary. And let’s not even get started on the spam volume in the old inbox. It was like a digital landfill!
We quickly realized we needed to create a 'Nothing' label to catch and ignore irrelevant emails. It was a lightbulb moment that saved us from drowning in junk drafts. Plus, we had to remove the old 'Handle' label after triage to prevent loops. Who knew email could be so complicated? But hey, that’s the beauty of discovery!
The real “ah-ha” moment came when we leaned into reuse and iteration instead of trying to reinvent the wheel. By prioritizing minimalism—keeping labels simple and intuitive—we made the system easier to debug and extend. It was like a breath of fresh air!
Adjusting the AI prompt to reinforce a first-person voice and emotional mirroring made the drafts feel personal and appropriate right from the first run. And when we hit a wall with Zapier tasks, we didn’t hesitate to upgrade our plan.
“It was bound to happen, but I ran out of zaps, so I had to sign up for an even MORE pro account...”
Honestly, we laughed through it. Sometimes you just have to roll with the punches!
Now that we’ve set up the draft-generation flow, the next steps are all about fine-tuning and monitoring. We’ll be keeping a close eye on the old inbox drafts for tone and completeness, and we’re excited to create Slack alerts for flagged investor action items. Plus, we’ll be working on labeling newsletters for subscription transfer clean-up.
We’re also planning to set up a mini dashboard to track old inbox clearance rates. Once everything stabilizes, we’ll dive into historical inbox processing in slow batches. It’s all about building a resilient foundation for long-term, low-maintenance processing.
Zapier: Workflow automation for email triage and drafts — https://zapier.com;
Gmail: Email organization and label management — https://mail.google.com;
Relevance AI: AI draft generation with custom logic prompts — https://relevanceai.com
Google Drive: Staging area for project documentation and checkpoint CSVs — https://drive.google.com
