1 of 24

Content Automation

Platform

An end-to-end AI-powered content production system

built on Airtable for a digital marketing agency

BEATRIZ BRAGA

Solo Builder

Airtable

Make

OpenAI / OpenRouter

Tally Forms

Bannerbear

2 of 24

01 — CONTEXT

The Business Problem

Marketing Agency / Content creation couldn’t scale with client growth

01 Manual & Repetitive Briefing

Manual onboarding for every client

No centralized source of truth

Time wasted re-entering the same information

02. Inconsistent Brand Voice

No standardized briefing system

Content varied across creators

Brand consistency was compromised

03 No Path to Scale

Growth required hiring more people

No operational leverage

Model was not scalable

→ Content production depended on people, not systems.

3 of 24

02 — THE SOLUTION

A centralized system to automate content creation

while preserving brand consistency

INPUT

Client Brief

Tally Form

Brand Voice

PROCESS

Airtable as the

heart of the system

Make.com

AI Generation

OUTPUT

Content Creation

Design Assets

Content recycling

From manual, inconsistent workflows → to a scalable, system-driven content engine

4 of 24

02 — THE SOLUTION

Built End-to-End

Elements created by me in collaboration with the operational team and leadership.

D

Data Architecture

Relational tables with linked records, lookup chains, and formula fields > the operational backbone of the system.

A

Airtable Build

All tables, views, field logic, and base configuration built from scratch, including formulas and automation triggers.

S

Automation Scripts

JavaScript scripts embedded in Airtable automations for webhook triggering and data extraction.

M

Make.com Scenarios

Scenarios orchestrating AI content generation and writing structured outputs back to Airtable dynamically.

I

Integrations

Tally, OpenAI via OpenRouter, and Bannerbear, external tools fully connected to the automated content pipeline.

5 of 24

Architecture Overview

TALLY

FORMS

Client intake

& profile quiz

AIRTABLE BASE

Clients

Profiles

Design Set

Planning

Prompt Library

Social Posts

Designs

Content Creator

MAKE.COM

Automation router

OPENAI

OPENROUTER

LLM content generation

BANNERBEAR

Design automation

Airtable

Make.com

AI / LLM

Design Automation

6 of 24

03 — CLIENT JOURNEY · STEP 1

Client Onboarding

BEFORE

2–3 onboarding calls required per client

Information scattered across notes and documents

No standardized briefing process

Rework needed for every new content request

AFTER

Automated onboarding

Centralized client data in Airtable

Standardized brand profiles for every client

Reusable context across all future content

HOW IT WORKS

1

Client completes a structured intake form (Tally)

2

Responses are scored and mapped to a brand profile

3

Data is processed and stored in Airtable

4

Client gets a persistent profile, history, and design system

↓ Onboarding time

from 3 meetings → 1 automated step

7 of 24

Airtable · Client Profile Record

8 of 24

03 — CLIENT JOURNEY · STEP 1

Structured brand data captured at onboarding, reused across every content generation request

EW

Dr. Emily Walsh

Aesthetic Medicine

PROFESSIONAL

Language

English (EN)

Platforms

Instagram · LinkedIn

Active since

January 2024

VOICE TONE

Authoritative · Data-backed · Results-oriented

FORMALITY LEVEL

High — Professional register

MAIN TOPICS

Skin Health · Anti-aging · Aesthetics · Wellness

TARGET AUDIENCE

Women 35–55 · Urban · Upper-middle class

EDITORIAL STYLE

Educational · Evidence-based · Long-form friendly

CALL-TO-ACTION

"Schedule your consultation"

TO AVOID

Slang · Competitor mentions · Aggressive promotions

CUSTOMER AVATAR

Career-driven · Invests in health · Researches before deciding

9 of 24

04 — CLIENT JOURNEY · STEP 2 & 3

Content Planning & Creation

From scattered, manual content creation → to an AI-driven, structured content pipeline

BEFORE

Content created without access to client history

No standard process — each writer worked differently

Hours lost searching past meeting notes and briefs

Scattered across multiple platforms and tools

AFTER

AI generates monthly plan from client's brand profile

Plan reviewed and approved before production starts

All posts created automatically, caption, date, visual direction

Team reviews and requests edits

HOW IT WORKS

1

Team selects client — brand data auto-fills from Airtable

2

One click triggers AI to generate full monthly content plan

3

Client reviews and approves the plan before production

4

Factory runs: all posts created automatically with captions + dates

↓ Content production time

from 5 hours per client → 1 hour

10 of 24

Airtable · Content Planning Pipeline

11 of 24

04 — CLIENT JOURNEY · OUTPUT EXAMPLE

Content Pipeline

Stage 1 generates the monthly plan — Stage 2 produces every individual post from it

STAGE 1

Monthly Planning

Client

Dr. Emily Walsh

Month

March 2026

Format

Instagram · Reels · Carousel

Total Posts

12

AI Status

✓ Done — Awaiting approval

CONTENT THEMES

Anti-aging skincare routines

Post-procedure recovery tips

Education: what is aesthetic medicine?

STAGE 2

Social Post Production

POST 03 · MARCH 7, 2026

"Skin doesn't lie. After just one microneedling session, the results speak for themselves — here's what the first 48h look like."

Visual direction:

Elegant close-up of skin texture · Cool tones · Minimal overlay text

Approved

POST 07 · MARCH 19, 2026

"The most common question I get: 'How early is too early to start aesthetic treatments?' The answer might surprise you."

Visual direction:

Professional headshot style · Warm lighting · Branded frame

Waiting Approval

12 of 24

Planning Table

Social Media Table

13 of 24

Prompt Library

Designs

14 of 24

05 — CLIENT JOURNEY · STEP 4

Content Repurposing

From one approved post → to content adapted for every channel and format

BEFORE

Newsletter, video script.. — each written from scratch

Same core idea reformatted 4× with no consistency across channels

Hours spent manually adapting content for each platform

Brand voice and tone drifted with every new format

AFTER

One approved post triggers repurposing automation in Make

AI adapts the message to newsletter, video script, and carousel

HOW IT WORKS

1

Approved social post triggers the repurposing automation

2

AI identifies the core idea, angle, and brand voice of the post

3

Format-specific versions generated: newsletter, video, carousel

4

All repurposed outputs land in Airtable, linked to source post

×4 Content output per approved post

15 of 24

Airtable · Content Repurposing Interface

16 of 24

Content Repurposing Creator

Prompt Library

17 of 24

Automation Scripts

Script 1 — Planning Trigger

TRIGGER

Record created in Planning — Content Type = "Planning"

ACTION

Sends RecordID to Make → LLM generates the full monthly content plan

Stable

Script 2 — Social Post Trigger

TRIGGER

Record created in Planning — Content Type = "Social Media"

ACTION

Sends RecordID + TableID + BaseID + OutputTableID → LLM generates a single post

Stable

↑ Same base trigger — filter distinguishes Planning vs. Social Media type

Script 3 — Output Normalizer

TRIGGER

When LLM output arrives from Make and is written to the Social Posts table

ACTION

Regex parses the raw LLM output and writes each piece of data to the correct Airtable field — maintaining data integrity across all content types

Unified

Replaces 2 separate extractors — one consistent logic for all content types

FIELDS EXTRACTED BY REGEX

Post Title

Name / identifier of the content piece

Publication Date

Scheduled date parsed from LLM output

Caption

Full caption text — multiline, preserved as-is

Client Name

Linked back to the correct client record

18 of 24

Airtable Automations and App Scripts

19 of 24

06 — INTEGRATIONS

The middleware that receives Airtable triggers, processes data, calls every tool in the stack, and writes structured results back

TRIGGER

Airtable

Record created or updated

TRIGGER TYPES

Planning record created

Social post record created

ORCHESTRATOR

Make.com

Automation platform — listens to Airtable, processes payloads, calls APIs, and writes structured results back

WHAT MAKE DOES IN EACH SCENARIO

Routes data between Airtable and any connected app

Parses, transforms, and filters each incoming payload

20 of 24

Make.com · Automation Scenarios

21 of 24

TEAM KANBAN

Airtable Interfaces

CLIENT DASHBOARD

22 of 24

RESULTS · MEASURED IMPACT

Concrete outcomes from implementing the full Content Automation system

1 step

was 3 separate meetings

ONBOARDING

Fully automated | zero human intervention required to bring a new client on board

1 hour

was ~5 hours per client

CONTENT PRODUCTION

Monthly content time reduced

15 clients

was 10 with the same team

TEAM CAPACITY

50% more accounts managed with no added headcount or working hours

1 database

did not exist before

CONTENT LIBRARY

A structured, searchable content base built from scratch — every post saved and reusable

23 of 24

CHALLENGES AND RESOLUTIONS

CHALLENGE

HOW IT WAS SOLVED

1

Mapping an unstructured operation into a clear process

No documented steps, no defined inputs or outputs anywhere

Deep-mapped every workflow stage, defined inputs and outputs for each step, then rebuilt the full operation as a structured Airtable base

2

Defining data standards so AI outputs were actually usable

No schema, no field conventions — AI returned freeform text

Established field schemas per table and ran iterative prompt refinements until outputs consistently matched the expected structure

3

Separating planning from content production

Both stages were treated as one blurred, unmanaged process

Split into two independent tables with separate automations

4

Getting consistent, structured output from the AI

Raw AI responses arrived in inconsistent formats — unusable as-is

Built the Output Normalizer automation to parse, clean, and route each field to the correct Airtable column automatically

5

Keeping human approval without creating a bottleneck

Manual review was essential but disrupted the automated flow

Kanban interface with defined stages (Editing → Waiting Approval → Approved) — one human touchpoint, zero disruption to the pipeline

24 of 24

Thank You

biabbraga@gmail.com

+55 81 991688491