The Problem

Modern advertisers face three key challenges:

1. Creative Production Is Slow

• Hiring creators takes time
• Product shipping delays production
• Coordination slows everything down
• Final delivery can take days

2. Scaling Creatives Is Difficult

• Every new ad requires a new production cycle
• Limited variations mean slower testing
• Slow testing leads to poor performance

3. Poor Attribution Breaks Optimization

• Duplicate tracking events
• Conflicting SDK data
• Inaccurate reporting

This leads to wasted ad spend and poor decisions.

The Challenge

We had to solve three complex problems at once.

1. Build a Multi-Model AI Video System

This wasn’t just about generating images.

We needed to generate realistic, conversion-focused video ads using multiple AI systems working together.

Challenges included:

• Combining outputs from different AI models
• Managing asynchronous processing
• Keeping generation time under one minute
• Maintaining visual and narrative consistency

2. Fix Broken Attribution

The system had:

• Duplicate install events
• Conflicting data sources
• Inflated metrics

This directly impacted campaign performance.

3. Keep the Product Simple

Even with complex AI behind it, the app had to be:

• Easy for non-technical users
• Fast and responsive
• Focused on results

Our Approach

We treated this as both a system architecture problem and a product design challenge.

Phase 1: Building the AI Creative Engine

We created a multi-model orchestration pipeline.

How It Works:

1. User uploads a product image
2. Selects from 300+ AI actors
3. Generates a video ad instantly

Behind the Scenes:

• Assets generated dynamically
• Video created using AI pipelines
• Outputs structured for ad platforms

This wasn’t a simple integration.

It was a production-grade AI system.

Phase 2: Fixing Attribution

We rebuilt the entire tracking system.

What We Did:

• Removed conflicting SDK signals
• Set a single source of truth
• Fixed event mapping for trials and purchases
• Enabled proper postback tracking

Result:

• Clean data
• Accurate reporting
• Better campaign decisions

Phase 3: Designing for Conversion

We simplified the product experience.

• Minimal steps
• Instant previews
• Clean UI
• Focus on output, not complexity

The goal was simple:

Reduce friction → increase output → improve performance

The Result

AI-Powered Creative Engine

• Generates video ads in minutes
• Produces lifelike UGC-style content
• Fully automated workflow

Scalable Creative Production

• Unlimited ad variations
• Faster testing cycles
• No production bottlenecks

Clean Attribution System

• No duplicate events
• Accurate tracking
• Reliable performance data

Business Impact

Up to 96% cost reduction vs traditional workflows
Faster iteration = better ROAS
Near-instant creative generation

Key Insight

AI alone doesn’t create value.

Orchestrated AI systems + clean data pipelines do.

Final Thoughts

This project was more than building an app.

It was about creating:

• A multi-model AI system
• A reliable data pipeline
• A scalable growth engine

At CIZO, we focus on building systems that don’t just work—they scale.

Want to build AI-powered systems like this? Let’s talk.