🎬 Content Production Pipeline

GB-CP · UpCounting · V0.5

In Development Owner: Abir (design, voice approval) Executor: Ilaria (day-to-day) Last updated: May 11, 2026

Content strategy (GB-CS) defines what to make and for whom. This document defines how to move from "a thing happened in a meeting" to "a post is live."

Core diagnosis: Abir's content gap is a retrieval problem, not an ideation problem. He performs well when prompted with specific source material; he stalls when starting from nothing. Production must therefore surface pre-vetted, pre-thought-through material at the moment of filming — not a creative session every time.

The 3-Stage Pipeline

Each stage is a separate Airtable table. Records flow forward; status fields gate the handoffs.

🌱
Stage 1 — Raw Extractions
tblVQHadtN6JdN24F
Compressed source material with full original context
→
💡
Stage 2 — Post Concepts
tblxAX5pLWWfNpw4u
Ready-to-shoot hooks with full scripts
→
🚀
Stage 3 — Posts
tblE2wYlewAcl1Me0
Filmed video + multi-platform copy variants
System of Record
Airtable — Content Library base
Shipping Gate
Every concept requires Pillar + CTA + Format
Current Status
Prototyped end-to-end; no shipped post yet

🌱 Stage 1 — Raw Extractions

Captures full context before any compression — what was said, by whom, why it matters. Preserves the names, numbers, and texture that make scripts credible.

Status Flow

⚡ New
Captured but not yet reviewed by Abir
Automated capture
↓ Abir reviews
📥 To Process
Approved for concept extraction
Abir
↓ Ilaria / agent drafts concept
✅ Done
One or more concepts drafted from this record
Ilaria / Agent

Sources of Raw Records

🎙️ TANA Voice Captures
Abir tags thoughts in TANA → local agent pushes to Airtable
Live · Automated
📊 Daily Strategy Review Agent
Reads Fireflies transcripts + Slack → writes content ideas to Raw table
Live · Automated
🎬 Meeting Clip Mining
Roundtables + Skool CFO office hours — automation planned
Manual today · Ilaria
🔁 Remixes
Re-mine best posts ~12 weeks post-ship via Metricool MCP
Planned
💡 Inspiration / Swipe File
External formats from TANA, adapted with our pillars + CTAs
Planned
🚗 Driving Voice Sessions
ChatGPT voice session with prompt questions → parsed into Raw table
Planned

💡 Stage 2 — Post Concepts

A specific, ready-to-shoot take derived from Raw records. Each concept = one argument, one hook, one full video script. When Abir sits down to film, he picks from a pre-vetted queue — not a blank page.

Status Flow

✍️ Drafted
Concept created; awaiting Abir's review
Ilaria / Agent
↓ Abir reviews — routes to one of three paths
🎥 Filmed
Abir filmed it
Abir
♻️ Edits Needed
Bones good; hook/script needs rework
Abir → Ilaria
❌ Killed
Rejected; kept for learning
Abir
↓ Post-produced and scheduled
🚀 Shipped
At least one linked Post is 🚀 Shipped
Ilaria

Required Fields per Concept

Hook (first 1–2 sentences)
Angle (thesis, 2–3 sentences)
Script (~45–60 sec)
Pillar (one of four)
Content type
Best Format
Strength (1–5)
CTA
Storytelling
Qualifier
Raw Extraction link
Notes

Content Pillars

💰 Ecom Finance
🔗 The Intersection
🏗️ Building the Machine
🧠 The Mindset

Building the Machine + The Mindset are organic-only pillars — CTA field may be None.

🚀 Stage 3 — Posts

A filmed video and its multi-platform copy variants. One row per filmed asset (Twitter/LinkedIn share long-form copy; IG gets its own caption field). Abir never touches this table — he ships by flipping a Concept to 🎥 Filmed.

Status Flow

📝 Draft
Ilaria filling video link, transcript, copy variants
Ilaria
↓ All variants approved + Scheduled Date set
📅 Ready to Post
Picked up by Metricool agent (GB-CP-04)
Ilaria → Agent
↓ Metricool schedules + writes back post IDs/URLs
🚀 Shipped
Post IDs written back to table
Metricool Agent
⚠️ Status flips on Metricool acceptance, not actual publish time. Use the 🔴 Is Live formula field for the real published-check.

Ilaria's Per-Concept Loop

  1. Post-produce video in Descript → save final cut + transcript to Google Drive
  2. Create Post row → paste Drive link + transcript → link to Source Concept
  3. Run GB-CP-03 prompt in Claude with transcript + Concept context → draft Long-form copy + IG caption
  4. Review/edit each variant → set Scheduled Date → flip to 📅 Ready to Post
  5. Metricool agent (GB-CP-04, planned) schedules across platforms → writes back post IDs → flips to 🚀 Shipped
  6. Once any linked Post is 🚀 Shipped → Concept also flips to 🚀 Shipped

Required Fields per Post

🎬 Video file (Drive link)
📜 Transcript (from Descript)
✍️ Long-form copy (Twitter/LI)
📷 IG caption
💡 Source Concept (link)
📅 Scheduled Date
📱 Platform (multi-select)
🎣 Primary CTA

Ownership at a Glance

Action Owner Trigger / Notes
Stage 1 — Raw Extractions
TANA thought → Raw table Agent Abir tags idea as thought in TANA
Daily Strategy Review → Raw table Agent Daily cron; reads Fireflies + Slack
Meeting clip mining → Raw table Ilaria Manual today; automation planned
Vet Raw records (⚡ New → 📥 To Process) Abir Abir reviews new queue
Stage 2 — Post Concepts
Draft concepts (Raw → ✍️ Drafted) Ilaria + Agent GB-CP-02 process doc in Notion Agent Processes DB
Approve / kill concepts Abir Routes to 🎥 Filmed, ♻️ Edits Needed, or ❌ Killed
Film (flip to 🎥 Filmed) Abir The only action Abir takes to trigger post-production
Stage 3 — Posts
Post-production + variant drafting Ilaria Descript → Claude (GB-CP-03) → Airtable Post row
Schedule across platforms Agent (GB-CP-04) Triggers on 📅 Ready to Post — planned

Pipeline Flow Diagram

End-to-end workflow from source capture to published post, across all three Airtable tables.

STAGE 1 · RAW EXTRACTIONS STAGE 2 · POST CONCEPTS STAGE 3 · POSTS 🤖 Sources TANA · Daily SR ⚡ New Captured, unreviewed 📥 To Process Approved by Abir ✅ Done Concept drafted Abir Ilaria / Agent concept drafted from raw record ✍️ Drafted Ilaria / Agent creates Abir routes 🎥 Filmed Abir shoots ♻️ Edits Needed Hook/script rework ❌ Killed Kept for learning Ilaria revises → back to draft post-production starts 📝 Draft Ilaria fills copy 📅 Ready to Post Ilaria approves 🚀 Shipped GB-CP-04 schedules Ilaria GB-CP-04 ↑ triggers Concept → 🚀 Shipped
Abir
Ilaria
Agent
Conditional / loop
Cross-stage transition

Tools & System of Record

📋
Airtable
System of record — Content Library base (3 tables)
🔥
Fireflies
Meeting capture → feeds Raw table via agents
🤖
Claude
Concept drafting (GB-CP-02) + copy variants (GB-CP-03)
🎵
Descript
Post-production + transcript generation
📅
Metricool
Multi-platform scheduling — GB-CP-04, planned
📝
Notion
This doc + GB-CS strategy + Agent Processes DB
⚠️ Connector caveat: Pipedream Airtable OAuth connector silently fails on schema changes and bulk writes. Agent-driven writes and schema changes currently use the Airtable REST API with a personal access token. To be fixed or replaced before this workflow scales.