In case you were curious, cutting review time by 95% is *checks notes* pretty good.
We cut manual review time of outgoing user emails by 95% with Zapier, OpenAI and HTML/CSS to Image! 📩 👇 We tried pricey 3rd-party email review options and got burned. Multiple phishing and emails escaped these automated filters, damaging RSVPify's email sender reputation for weeks, and leaving us scrambling. As a freemium product operating in an increasingly stringent email delivery environment, we were forced to begin manually (and painfully) reviewing each and every outgoing email sent by our non-paying and untrusted new users. Thousands of emails a week. Until we tried something new using a series of "Zaps": 1. Query the database for new emails pending review 2. Run an AI-generated Javascript code to extract and categorize email variables like merge tags, logos, images, buttons, links and CSS 3. Use HTML/CSS To Image to generate an image file visualizing the outbound email, including embedded images 4. Run the image file through ChatGPT Vision with a carefully scripted prompt to detect SPAM, phishing, hate, adult content and other sensitive non-event related emails 5. Log email approvals and blocks in Slack 6. Send our systems a webhook when an email is approved, and automatically move email into delivery queue A couple months later, we've cut manual email review time by upwards of 95% -- and now we manually review only those emails which AI could not confidently approve!