See llms.txt for all machine-readable content.

Back to Templates

Screen podcast guest pitches in Gmail and draft replies with GPT-4.1-mini

Created by

Created by: ReadyUpAgents || readyupagents
ReadyUpAgents

Last update

Last update 2 days ago

Categories

Quick Overview

This workflow monitors Gmail for new emails, screens potential podcast guest pitches with rule-based spam checks and OpenAI classification, then applies Gmail labels and optionally drafts personalized replies, plus a daily digest email summary.

How it works

  1. Polls Gmail every minute for new emails and pulls message content.
  2. Normalizes each email (sender, subject, body, links) and fetches message metadata from the Gmail API to list attachment names.
  3. Scores spam and scam signals using rule-based checks on sender details, links, and attachment names.
  4. Uses OpenAI to classify whether the email is a guest pitch and to rate topic match, fit score, and risk based on your show settings.
  5. Combines the rule score and OpenAI output into a verdict (Strong Fit, Maybe, Not a Fit, Spam/Pay-to-Play, or Not a Pitch) and creates any missing Gmail labels before applying the correct label to the message.
  6. For Strong Fit and Maybe pitches, uses OpenAI to write a reply draft, skips threads that already contain a draft, and saves the new draft in Gmail with placeholders highlighted and an optional “Needs Details” label when required.
  7. On a daily schedule, compiles the saved screening results into a text digest and emails it to the configured owner address.

Setup

  1. Connect a Gmail OAuth2 credential for all Gmail nodes, including the HTTP request to the Gmail API and the steps that create labels, apply labels, create drafts, and send the digest.
  2. Add an OpenAI credential for the two OpenAI steps (classification and reply drafting) and confirm the model ID in Settings (default: gpt-4.1-mini).
  3. Fill in the show profile in Settings (show name/about, topics, optional audience/must-haves/dealbreakers, reply style, booking link, and scoring thresholds) and optionally add trusted_senders/known_bad_senders.
  4. Decide which drafts to create by toggling draft_strong_fit and draft_maybe, and set the Gmail label prefix if you want something other than AI/.
  5. If you want the daily digest, set Digest Settings owner_email (and optionally send_when_empty) and confirm the schedule time matches your instance timezone.