Streamlining Your Workflow: Gmail Labels and AI Drafting Made Easy!

Summary:

This checkpoint centered around building a full triage-to-draft flow: automatically labeling Gmail messages and passing the ones requiring a response into Relevance AI for a context-aware, emotionally intelligent draft. While the flow started simple, the journey into label IDs and trigger logic revealed a mess of Zapier-Gmail weirdness, prompting a creative workaround involving a lookup table to map human-friendly labels to Gmail's required internal label IDs.

How Can We Tame the Inbox Beast?

Ah, the inbox. It’s like a never-ending firehose of information, and if you’re anything like me, it can feel overwhelming. Our latest checkpoint was all about transforming that chaotic stream into a manageable flow. We set out to build a triage-to-draft system that would automatically label incoming Gmail messages and identify which ones needed a response. The goal? To reduce the overwhelm and make our inbox feel less like a battleground and more like a conversation.But let me tell you, the journey was anything but smooth. We quickly discovered that Gmail’s labeling system is a bit of a labyrinth. Instead of straightforward names, we were faced with cryptic internal label IDs. I mean, come on, Google! Who thought that was a good idea? As we dove into the depths of Zapier and Gmail, I found myself exclaiming,

“this is literally insane!!!”

when I realized Zapier couldn’t even list all the labels for us. Talk about a buzzkill!

What’s the Secret Sauce for Smart Email Labeling?

The heart of our project was to create a structured labeling system that could intelligently categorize emails. We envisioned categories like Respond, Work, Support, and Revenue, each with its own subcategories. But as we started building, we hit a wall. The idea of manually mapping every label to its internal ID felt like a Sisyphean task. I remember staring at my screen and thinking,

“ok, but do I really have to manually map everything???”

That’s when the lightbulb moment struck! Instead of drowning in a sea of manual data entry, we decided to create a lookup table. This table would match user-friendly labels to Gmail’s internal IDs, allowing us to automate the process. It was a game-changer. When we finally got it working, I couldn’t help but shout,

“HOOOOOLYSHIT… it did work... so now I’ll do all that data entry!”

How Did We Make Email Drafting Feel More Human?

Once we had our labeling system in place, it was time to tackle the drafting process. We wanted to ensure that the emails generated by Relevance AI reflected Danielle’s voice and editorial tone. This was crucial for maintaining a personal touch in our communications. We spent hours refining the drafting prompt, making sure it was clear, flexible, and emotionally intelligent. The result? A system that not only filtered out the noise but also crafted responses that felt genuine. We were able to create drafts with optional tone paths—accept, decline, clarify—making it easier to respond appropriately. It was like giving our emails a personality! But let’s be real, the road to this point was littered with frustrations. I distinctly remember saying,

“I just want to go on record to say this is dumb”

as we navigated the confusing world of nested labels. It felt like we were trying to solve a Rubik’s Cube blindfolded.

What Surprised Us About This Process?

The biggest surprise? The power of slowing down to acknowledge what wasn’t working. Instead of rushing to find a solution, we took a step back and realized that the label mapping was the real constraint. Once we identified that, everything clicked into place. This phase was all about *infrastructure with personality*—creating emotionally aware automation rather than robotic filtering.The combination of Relevance AI’s tone-matching draft generator and our label filtering made the process feel like a dialogue again. We transformed our inbox from a chaotic firehose into a structured conversation. The formatter lookup table became our unsung hero, removing complexity and opening doors for other systems.

What’s Next on Our Journey?

As we look ahead, we’re excited to maintain and expand our Gmail label map and lookup table. We plan to add new prompts for different types of emails—think podcasts, partnerships, and even friendly catch-ups. The goal is to continue refining Relevance AI’s tone for various response types and automate the movement of completed emails into folders. We’re also exploring a parallel flow for content generation from labeled newsletters. The possibilities feel endless, and we’re just getting started! In closing, this journey has been a wild ride filled with unexpected challenges and delightful discoveries. We’re building something truly special, and I can’t wait to share more updates as we continue to refine our systems. Here’s to creating a more connected, emotionally intelligent world—one email at a time!

Tools we used:

Zapier — https://zapier.comGmail — https://mail.google.comRelevance AI — https://relevanceai.comSlack — https://slack.comGoogle Sheets — https://sheets.google.com