Grace Clarke's Claude Workflows for Business Automation and Rebuilding Gmail

In this episode of How I AI, I was so excited to sit down with Grace Clarke, an AI teacher and former marketing consultant. One of my core beliefs about AI’s impact is that it will allow all of us to massively differentiate on the quality of our service. The bar for human relationships and customized client work is about to get much higher, because AI gives us the leverage to do more and make everything more personal.

Grace is living proof of this idea. She runs a relationship-driven, process-driven business and found herself drowning in administrative work—a pain I think many of us can relate to. Instead of hiring an assistant, she decided to teach herself how to use Claude as a true collaborator. She has completely transformed her business by building automated systems that not only save her time but also deliver a beautiful, high-touch experience for her clients.

We get into two incredible workflows she built from the ground up. The first is her "Pipeline Operator," a fully automated system that generates stunning, interactive client proposals and onboarding materials. The second is a project born out of pure frustration: a complete recreation of Gmail that lives inside Claude, allowing her to manage email without ever opening the dreaded Gmail tab. Grace’s approach is a powerful reminder that you don't need to be a 'technical' person to build amazing things with AI; you just need a problem to solve and a willingness to collaborate with the tool.

Workflow 1: The Automated Client Proposal Pipeline

Grace was spending upwards of 20 hours a week on administrative tasks, which was keeping her from the actual teaching and consulting work she loved. She faced a deluge of emails, found the client relations process devoid of emotion, and was constantly juggling 20 open tabs. Her solution was to build a pipeline operator in Claude that wakes up every hour to manage her client workflow.

This system produces beautiful, personalized client-facing materials that immediately set the tone for the work they’ll do together. It’s an advertisement for the power of AI from the very first touchpoint.

Step 1: Intent Engineering, Not Prompt Engineering

Grace has a strong take: "prompt engineering is dead, but intent engineering is where we need to be focusing our time." Instead of crafting a perfect, multi-page prompt, she started by simply talking.

Using the Claude mobile app on a walk, she spoke for two or three minutes, laying out her problem and what she thought she wanted. She put the pressure on Claude to come back with a solution. Her core instruction was a conversation, not a command.

Claude came back to me and said, I think we need to be creating an interactive artifact that's password protected. What do you think? And we went back and forth. That is a prompt going from a hyper-engineered chunk of text to a conversation.

This conversational process, which took about 10 minutes of talking and an hour of iteration, formed the foundation for the entire workflow.

Step 2: Building the Core Skill Files

The pipeline runs on a few key skill files. The two most important are the proposal_maker and her voice_guide.

• The Proposal Maker: This skill file is the blueprint. It contains the rules, steps, and structure for every proposal. Grace versions her files with naming conventions, which helps her feel more connected to the technology she’s building. The skill outlines her teaching philosophy, confirms the deal shape, and lays out the document structure.

• The Voice Guide: This is more than just a style guide; it's a "think like me" guide. It started with a simple instruction to Claude: "I think I need a guide that will make everything more like me." It now contains her communication philosophy and, most importantly, a list of words and phrases to avoid—the AI slop we all see on LinkedIn. She constantly updates it by voice-noting examples of what she doesn't want to sound like.

Step 3: Generating Interactive and Automated Assets

Once the skills are in place, the pipeline automatically generates a suite of assets for each new client:

1. Interactive HTML Proposal: A beautiful, branded, and password-protected document hosted on Netlify. It reflects all the context from previous conversations.
2. Automated Pre-work: The system pulls updated documentation from AI labs and creates an interactive pre-work site for clients. It even tracks their progress and nudges them if they fall behind.
3. Pre-Session Questionnaires: The skill generates a custom questionnaire (built on a Google Form but styled to match her brand) to gather themes and understand where a team is stuck before a session.

This whole process turns a standard, boring proposal into a 'wow' moment that also serves as a vision for what the client will be able to build themselves.

Workflow 2: Rebuilding Gmail Inside Claude

Like many of us, Grace hates email. She describes Gmail as a place where "everybody has a key to your front door and can come leave things in your house." Staying in Gmail also means your work and replies are locked away; the learning doesn't compound. So, she decided to rebuild her email experience inside Claude.

She hasn't intentionally opened Gmail in a month. This project is a perfect example of something that feels highly technical but is achievable for anyone in about half an hour.

Step 1: Start with a Rant in Claude Code

Grace kicked off this project in Claude Code, which she finds faster and more proactive for ambiguous or technical-feeling projects. Her prompt was simple and emotional: a rant about how much she hates Gmail and never wants to be in it again.

From there, she and Claude went back and forth. It mocked up an interface, pulled in real information, and they iterated on the design. Could it add a link? Could it add bolding? Could it push a draft to Gmail?

Step 2: Create Custom Connectors for Full Integration

To make this work, Claude needed access to other services. Grace found that Claude struggled to write directly to Google Sheets or Google Docs. This led her to a crucial learning step: creating custom plugins.

This process involved setting up a Google Cloud project and a service account, which taught her about scoping and permissions. These custom connectors now live in her Claude settings, giving her model the specific access it needs to execute tasks.

Step 3: Move from Code to Co-work for the UI

Once the backend logic and connectors were scoped out in Claude Code, she wanted a more hospitable UI for the visual part of the project. She simply asked her Claude Code session to create a markdown file for the handoff.

"Write me a markdown file. And Claude saved a markdown file, a session handoff to my desktop. And I went right back into co-work, opened up a new session, and did nothing other than drag that markdown file in here. And co-work took over from there."

Now, she operates entirely within this custom interface inside Claude Co-work. She can read, draft replies, and with the click of a button, push the final email as a draft into Gmail, which pops open in a new browser window for a final send. It's a powerful way to triage email while training her AI on every single interaction.

Lightning Round Hack: The Automated Workout Tracker

Beyond these big business projects, Grace uses AI for tiny, personal life hacks. One of my favorites she showed me was her automated workout tracker. She likes to keep a record of her workouts but doesn't want the hassle of manually entering them into a spreadsheet.

Her process is brilliantly simple:

1. After a workout, she opens the Claude mobile app.
2. She voice notes what she did, for example: "I worked out today, did Bulgarian split squats at the gym in Nantucket. Add to tracker."
3. Claude parses the information and automatically updates her workout tracking spreadsheet.

She never has to look at the spreadsheet. She just trusts Claude to put the information where it needs to go. It's a perfect illustration of letting go and letting the AI handle the mundane tasks of life.

The Muscle Memory of AI

Grace’s workflows are inspiring, but her thoughts on AI adoption are just as important. The biggest hurdle for most people isn't the technology—it's building the muscle memory to turn to AI in the first place. My favorite piece of advice from our chat was her concept of a "forcing function."

"Set a Slack reminder or a Google Calendar alert that says whatever you're doing, screenshot it and put it into Claude. Get the whole window and just ask Claude, 'Could you help me with this?' We are just trying to build the muscle of deferring to Claude."

It’s not about having the perfect prompt or a grand project. It’s about creating a habit. By starting with a screenshot and a simple question, you show people the magic of inference and build their confidence. These tools are here to collaborate with us, not just take orders. As Grace demonstrated, if you start with your problem and have a conversation, you can build incredible things you never thought possible.

Episode Links

• Guest: Grace Clarke
• Host: Claire Vo