← Back to selected work

Case study

Mindvalley Email Builder

An AI agent that turns email briefs into published, production-ready templates.

Context

Marketing and lifecycle teams write email briefs as Google Docs — copy, links, images, targeting notes. Turning that into a live, on-brand, technically correct email template in the company's email platform was a separate, manual step every time.

The problem

The manual version of this job wasn't hand-coded HTML — it was the email platform's own drag-and-drop editor, used the slow way. Someone had to copy the brief's content out of Google Docs, paste it into the editor, then manually redo all the text formatting on top, add every hyperlink one at a time, and handle every image separately — downsizing it first, then importing it into the platform before it could even go into the email. It's a lot of repetitive manual assembly for something that should be mechanical, and each of those steps is a chance for something to slip — a paragraph left unformatted, a link missed, an image handled differently than the last one.

What I built

An AI agent that takes a brief — a Google Doc link or pasted copy — and turns it into a published, production-ready email template with no manual coding step:

  1. Reads the brief directly from Google Drive (or from pasted text)
  2. Infers the email type from content across a defined set of categories
  3. Scans the entire document for every real hyperlink — and every "look-alike" link (styled like one but missing a URL) — and confirms them before building, rather than guessing or inventing CTAs
  4. Automatically detects images referenced in the brief, downloads them, optimizes them for file size, imports them directly to the email platform, and inserts the resulting hosted image URL into the template — no manual upload or linking step
  5. Maps the intended audience to the correct legal disclaimer block automatically
  6. Generates the full HTML and runs itself through a fixed technical checklist before ever showing it — correct button markup, correct border-radius per card position, correct footer pairing for the email's type and language, exactly one unsubscribe link, correct templating syntax — fixing anything that fails before output
  7. Assembles the internal naming convention from the inferred and confirmed fields
  8. Publishes directly to the email platform via API, updating the template if one already exists, and flags a test-send as the final human check

The workflow is fixed and repeatable by design — same steps, same questions, same validation, every time — so it doesn't drift the way ad-hoc manual builds do.

Outcome

Cuts email production time by roughly 70%, from brief to published, technically-validated template.

Stack

Google Docs/Drive LLM agent (fixed workflow) Braze (email platform + API)