AI-FirstCreativeProductionWorkflow

+60%

Production capacity increase

without adding headcount

Impact

+60%

Production capacity

Without adding a single headcount

3–4 days faster

Avg. delivery reduction (2025)

Structured intake and auto-prioritization cut delivery from ~5 days to ~1 day

Same-day

Delivery time (2026 update)

AI platform: brief in → assets out, same day

→ 0

Brand & copy errors

Previously 1–3 spelling or tone human errors per month reached final delivery

My team went from managing creative requests by email to running a system anyone at Actinver can use.

Creative Ops — Before vs After
From reactive inbox to a structured production system
Before
01Requests via email
No structure, no priority
02Manual triage
By one person, urgent invisible
03Incomplete briefs
1–3 days back-and-forth
04Manual delegation
No ownership tracking
05No QA checkpoint
Errors reach delivery
065–8 days average
No visibility for executives
After
01Intake form
One entry point
02Auto-assignment
By type + complexity
03Structured brief
Complete from the start
04Stage-based tracking
With ownership
05AI validation layer
95%+ error reduction
062–4 days faster
Executive dashboard live

Context

Company

Actinver — national investment bank with multiple business units generating continuous demand for audiovisual content: social media, CRM, DOOH, video, events, paid campaigns.

Timeline

2024 – 2026 — initial system shipped and iterated; 2026 AI platform expansion

My Role

Nobody asked for this.

I mapped the full creative process, found where it broke, designed the fix, built it in Airtable, and trained the team.

Then I pitched it to executives as a performance-visibility tool — that changed how Marketing was seen internally.

The Problem

I mapped the full creative process. Every request ended up in one person's email. The deeper issue was organizational: Marketing's output was invisible to executives because there was no system to make it visible.

No Structure

Everything arrived through one inbox

Urgent and non-urgent requests competed invisibly

Incomplete briefs arrived constantly — 1–3 days lost to back-and-forth

Manual triage by Design Lead every single morning

No Visibility

Nobody could see the work or measure it

No backlog — other teams couldn't track their requests

Marketing's output was invisible to executives

No metrics, no throughput data, no capacity signal

No Capacity

The team was always reactive

No quality checkpoint — errors reached final delivery

After-hours work became routine, not exception

Volume grew but the system never scaled with it

The Solution

A creative operations system built on Airtable, supported by Slack and Institutional Mailing. Intake to delivery — structured, automated, visible.

Production process

Intake & Request

AirtableSlack

The requesting team fills out a structured form with objective, audience, channels, and deadline. Marketing receives an automatic alert. If the piece involves client communication, it goes through CRM approval before production begins.

Structured briefAuto-confirmationKickoff if neededUrgency flag

Brief & Kickoff

TeamsMeeting

For complex projects, an alignment meeting is scheduled with all involved parties — covering key messages, legal requirements, reference materials, and committed delivery dates.

Clear objectivesDefined audienceConfirmed channelsLegal inputs

Content & Design

CopyDesignAI

Content writes the narrative and sends it for sign-off. Once the copy is approved, Design begins. AI assists with visual variations, tone review, and iteration speed across formats.

Copy → Sign-offRevisions if neededCopy done → Design startsAssets uploaded

03.1

AI Generative Platform2026 Update*

Claude CodeNanoBananaStandalone

Once a request is approved, the platform picks it up automatically. It reads the brief, pulls from NanoBanana Pro and free stock, generates the piece grounded in the Actinver brandbook, adapts it for every social media format, and delivers it back to Airtable — reviewed for visual quality and copy accuracy. Some pieces that used to take a full day now take less than a minute.

Airtable triggerBrandbook-groundedAuto-adapt formatsVisual + copy QA

QA & Validation

AI ReviewGO / NO-GO

Internal review: the responsible person and Content, Design, and Social leads validate each piece. An AI-assisted checklist runs before final approval. Only a GO clears the piece for delivery.

Strategic alignmentBrand toneArt & identityCTA + accessibility

Delivery & Archive

Email48 hrs

The requester receives a link to their materials. They have 48 hours for minor feedback or final approval. Once the window closes, the request is automatically archived with no further changes.

Delivery link48h feedback windowAuto-archivedNo changes after close

* Step 03.1 was added in 2026 as a standalone AI generative platform, built entirely with Claude Code. Details in "The Platform" section below.

Execution

I built the entire Airtable architecture and automation logic, then ran two tracks.

Internal: onboarding for Marketing and Growth, company-wide docs, and two iterations — v1 was intake and assignment; v2 added the AI validation layer and the executive dashboard, once it was clear the real gap wasn't speed. It was visibility.

Organizational: I pitched the system to executives as a performance-visibility tool — shifting Marketing from a hard-to-measure service to a structured operation with metrics.

We treated it like a product: versioned, iterated, improved on real feedback.

System Rollout
Treated as a product — built, piloted, iterated
01Build
Airtable architecture · automation logic · intake form design
02Pilot
Small group test · Marketing + Growth · feedback collection
03Onboard
Training sessions · documentation shared · cross-company
04Iterate
Round 1 structural fixes · Round 2 UX improvements from real usage
05System live
Adopted by Marketing, Ops, and Digital Lab

Key Deliverables

Centralized intake form

Single entry point

Eliminated incomplete briefs and all email requests — no form, no request.

Automated assignment workflow

Stage-based tracking · Ownership · Deadlines

Removed manual delegation. Every request routed by type and complexity.

AI validation layer

Automated QA before delivery

Brand and copy errors dropped to zero — previously 1–3 spelling or tone errors per month reached final delivery.

Executive visibility dashboard

Real-time backlog · Production metrics

For the first time, Marketing's output was legible to leadership.

Documentation & training program

Company-wide adoption · v1 → v2 iterations

Ensured the system lived beyond its creator — and other teams could replicate it.

AI social media creative platform

2026 Update · Social Media Production

Standalone AI platform that reads the brief from Airtable, generates the asset grounded in the brandbook, adapts it for every format, and delivers it back — reviewed and ready.

Outcome

First two months: production capacity up 60%, delivery 3–4 days faster, after-hours overload gone.

Operations and Digital Lab replicated the system for their own teams — no training, no coordination from me. That's the best validation a system can get.

It also gave the team something it never had: data to prove Marketing's impact — evidence, not perception.

What the system made visible

What was being produced

Full inventory of assets by type, volume, and requester

For which channels

Social, CRM, DOOH, video, events — all tracked per request

In which formats

Format-level breakdown enabling smarter resource planning

At what volume

Hard numbers to justify headcount, tools, and budget decisions

The Platform

2026 update

The ops system solved the process. To go faster, I built the next layer — a standalone AI-powered creative platform.

Prototyped in Figma Make to validate the concept, then migrated to Claude Code — from prototype to production system.

When Airtable registers a new approved request, the platform picks it up automatically: reads the brief, pulls from NanoBanana Pro and Actinver's photo stock, generates the piece grounded in the brandbook, adapts it for every social format, and delivers it back to Airtable — reviewed for visual quality and copy accuracy. Same-day turnaround.

Before

1–2 days

Average time to produce a standard content piece

After — Platform

< 10 min

Triggered, generated, reviewed, and delivered automatically

Standalone

Any team member can use it independently

The goal wasn't to create a tool only I could operate. It was to build something the whole team could own — without needing a designer in the loop.

Built with AI

Entirely built with Claude Code

A solo build. No engineering team, no agency. One person, one AI tool, one production system in less than two months.

Executive backing

Endorsed by Actinver's main executives

Recognized as a concrete example of how AI can be implemented as a primary tool — not a novelty, not a test, but a daily production system.

Reflection

The team already had the capacity — it just wasn't visible or organized. The right process unlocked more output than headcount would have. Every asset was designed by a human; AI's job was to catch what humans miss at scale — errors, tone drift, brief misalignment — before anything reached the client. Not replacing judgment. Extending it.