Case Study

Viral Hook Library

An agentic AI workflow that architects a self-building viral hook library in Airtable. It automatically searches, extracts, and organizes content daily, with zero manual input.

Part One

Claude AI Agent

Objective

To build an end-to-end AI agent that reads from a UGC viral content newsletter from Gmail, extracts hooks and stores them in a structured Airtable database that builds itself daily.

Pipeline at a Glance

Viral hook agent pipeline — Gmail to Claude to Airtable, runs daily

How the Agent Works

  1. 1
    Connect to GmailAuthenticates via OAuth2; headless after first run.
  2. 2
    Query only new emailsReads last email received from a specific email newsletter.
  3. 3
    Deduplicate against AirtableGmail Message IDs prevent duplicate rows.
  4. 4
    Strip email noiseRemoves footers, ad copy, unsubscribe links pre-AI from the email body.
  5. 5
    Extract hooks with Claude AIReturns a clean list of hooks only.
  6. 6
    Save to AirtableLogs title, date, hooks as multiselect tags + raw text.
  7. 7
    Update run historyWrites today's date on success; prevents silent skips.
  8. 8
    Repeat dailySelf-schedules every 24 hours, no manual input.

Rules Given to the Agent

  • Extract only viral hooks. It doesn't extract opinions, email ad copy, or CTAs.
  • Return row values on Airtable.
  • If no hooks found, leave empty.
  • Use Gmail Message IDs to avoid duplicates.

Tools & Stack

ToolRole
Gmail API (OAuth2)Authenticate and fetch from target sender
Claude AI — AnthropicExtract viral hooks from cleaned email bodies
AirtableStore hook data, and deduplication layer
Python + scheduleRun pipeline daily self-contained

Results

22
Emails processed in first run
0
Duplicates across multiple runs
< 2 min
Full pipeline runtime end-to-end
Daily
Automated, zero manual input after setup
Part Two

Airtable Field Agent

The Airtable hook library feeds a second AI layer. An Airtable Field Agent reads incoming hooks daily and rewrites them in the voice of Chess BFF, a beginner-friendly chess app built around fun, rivalry, and chaos.

Objective

To build infrastructure that continuously analyzes content inputs and turns them into brand-ready outputs. It does everything with the voice of Chess BFF in mind and with zero manual rewriting.

Pipeline at a Glance

Field Agent pipeline — Airtable hook library to Chess BFF brand hooks

How the Field Agent Works

  1. 1Reads viral hooks sourced daily by the pipeline agent.
  2. 2Rewrites each hook in Chess BFF's brand voice.
  3. 3Generates 10 brand-native hooks per session.
  4. 4Produces villain arc, couples, and best-friend rivalry variants.
  5. 5Optimizes all hooks for TikTok, Reels, and Shorts under 15 words.
  6. 6Outputs map directly into the social content calendar.

Results

A fully automated content intelligence stack. It goes from newsletter inbox to brand-ready TikTok hooks with zero manual steps in between. It sources, extracts, organizes, and rewrites to brand guidelines all on their own, on a daily basis.

Airtable Base Preview

Airtable base grid view — Hooks from Gmail
Airtable base detail view — Chess BFF Viral UGC Hooks