TL;DR: This article details the “Built This UGC” system—an autonomous AI agent that creates high-performing user-generated content (UGC) ads from scratch, including market research, scripting, video generation, and subtitles—based on a proven $2.
📋 Table of Contents
Jump to any section (19 sections available)
📹 Watch the Complete Video Tutorial
📺 Title: I Built This AI UGC Ad Agent From a $2.5M Playbook
⏱️ Duration: 886
👤 Channel: Arseny Shatokhin
🎯 Topic: Built This Ugc
💡 This comprehensive article is based on the tutorial above. Watch the video for visual demonstrations and detailed explanations.
If you’re running an e-commerce brand or digital agency, you don’t need another operations hire—you need AI employees that do reporting and ad creation autonomously. In this comprehensive guide, we’ll unpack every detail from the groundbreaking “Built This UGC” system: a fully autonomous AI agent that generates high-performing, ready-to-use UGC (User-Generated Content) ads from scratch—market research, scripting, video generation, subtitles, and all.
Based on a real-world process that generated over $2.5 million in revenue using Sora 2 ads, this open-source AI agency doesn’t just automate—it collaborates, iterates, and refines like a human creative team. And best of all? You can deploy it for free in under 60 seconds.
Below, we break down everything you need to know: how it works, step-by-step setup, real examples, cost analysis, limitations, training methodology, and how to customize or resell it for your own niche.
Why You Don’t Need Another Ops Hire—You Need AI Employees
The core premise of “Built This UGC” is simple yet revolutionary: instead of hiring more staff to handle ad creation, reporting, and creative iteration, deploy AI agents that act as full-fledged employees. These agents don’t just output static content—they engage in dialogue, accept feedback, and refine results until they’re perfect for your brand.
As the transcript states: “You don’t need another ops hire. You need AI employees that do reporting.” This system is designed to replace entire workflows, not just speed them up.
Free 7-Day Proof of Concept—Risk-Free Trial
Before committing, you can test the system with a free 7-day proof of concept. The offer is bold: “It doesn’t clearly pay for itself? You keep the agent and pay nothing.” This underscores the confidence in the system’s ROI—especially for e-commerce brands running performance ads.
Real Example: 100% AI-Generated UGC Ad for Herbalux
The video demonstrates a fully AI-generated ad for “Herbalux,” an organic, fragrance-free aloe moisturizer. The ad script:
“If your gentle moisturizer still burns or turns your skin red, your damaged barrier is reacting to hidden irritants. That’s why I switched to Herbalux… If your skin freaks out easily, tap to try Herbalux.”
Critically, everything—from market research to scripting to video generation—was done autonomously by an AI agent. No human intervention was needed until final review and refinement.
How the AI UGC Ad Factory Works: A 3-Step Agent Workflow
The system uses a multi-agent architecture, where specialized AI “employees” hand off work sequentially. Here’s the exact flow:
Step 1: Strategy Agent – Market Research & Foundational Documents
After you answer a few questions about your product, the Strategy Agent generates four foundational market research documents:
- Product Document – Detailed brand and product description
- Offer Brief Template – Based on proven e-com agency frameworks
- Avatar Sheet – Ideal customer profile with beliefs and pain points
- Competitor Analysis – Insights from live market research
The agent also performs web searches and analyzes your website to build accurate brand context.
Step 2: Brand Agent – Storyboard & Script Creation
Once you approve the foundational docs, you’re transferred to the Brand Agent, which generates a complete storyboard with scene descriptions and scripts. You can request changes—like making it funnier or shorter—before proceeding.
Step 3: UGC Agent – Video Generation, Editing & Final Output
The final UGC Agent:
- Generates product images using AI (e.g., Nano Banana/Gemini 3 Pro)
- Creates detailed video prompts for Sora 2
- Generates multiple video clips in parallel
- Stitches clips together
- Adds subtitles
- Organizes all assets into folders like a real video editor
The result? A ready-to-use, platform-optimized UGC ad—no post-production needed.
What Makes “Built This UGC” Different from Other AI Ad Tools
Most AI ad generators are one-shot automations. This system is different because:
- ✅ Trained on a real $2.5M e-com process—not theoretical workflows
- ✅ Supports iterative refinement—you can ask to replace scenes, remix videos, or adjust tone
- ✅ Uses persistent file storage—assets and documents are saved across sessions
- ✅ Organizes output like a human editor—folders, versions, and metadata included
- ✅ Integrates real-time brand assets—logos, colors, and product images are applied automatically
Step-by-Step: Deploy Your Own UGC Ad Factory in 60 Seconds
Here’s how to launch your own AI ad agency for free:
Prerequisites
- Access to Agency AI platform (free tier available)
- OpenAI API key (for Sora 2 video generation)
- Gemini API key (for image generation)
Deployment Steps
- Go to Agency AI platform and navigate to the Marketplace
- Select the UGC Ad Factory agent
- Choose your AI models:
- Video Model: Sora 2 (standard) or Sora 2 Pro (premium)
- Image Model: Gemini 3 Pro (aka Nano Banana)
- Upload your brand logo
- Add your API keys:
- OpenAI:
platform.openai.com → Dashboard → API Keys - Gemini:
Google AI Studio → Get API Key (bottom left)
- OpenAI:
- Click Save—your agency deploys in ~60 seconds
- Enable Persistent File Storage (critical for asset reuse)
- Deploy as a Custom GPT for a whitelabeled web app experience
Real Demonstration: Creating a Hot Sauce UGC Ad
The transcript walks through a live demo for a fictional product: “Spicy Prompt” hot sauce, sent to clients as a post-engagement gift.
Input Details Provided
- Product Name: Spicy Prompt
- Heat Level: “Incredibly spicy—the spiciest you’ve ever tried”
- Role in Business: Client gift after project completion
- Target Audience: Marketing agency owners making $500K+
- Brand Voice: “Playful but sharp”
- Website: Attached for AI research
💡 Pro Tip: Always attach your website—“this agent can also perform research on exactly what it is that your company [is] doing on your avatar.”
Output Highlights
The final ad script:
“Most agencies send a polite thank you email. We send a bottle of hot sauce so aggressive it should come with HR training… One tiny drop on your taco feels exactly like opening Slack and seeing the words ‘quick question.’ Your brain just goes, ‘Oh no.’”
The agent generated:
- Two branded product image variants (yellow/blue colors, “AI infused” label)
- Three 30-second video clips using Sora 2
- Final stitched video with subtitles and logo overlay
Iterative Refinement: Fixing and Improving AI Output
AI video models rarely get it perfect on the first try. The power of this system lies in human-in-the-loop refinement:
- When the second clip had a cut-off script, the user asked the agent to fix the timing
- Requested to add the logo to the hot sauce bottle
- Asked to trim one second from the end of the third clip
Each request was handled seamlessly because the agent retains context and assets—“you won’t ever get this right on the first attempt… you actually have to iterate with the agent.”
Cost Breakdown: Is AI Video Generation Worth It?
Yes—but be aware of token costs. For the hot sauce ad:
| Cost Component | Amount | Notes |
|---|---|---|
| Total Cost | $9.00 | For a full, production-ready UGC ad |
| Sora 2 Video Generation | $8.20 | 90%+ of total cost |
| Sora 2 Pricing | $0.10/second | Standard model |
| Sora 2 Pro Pricing | $0.30/second | Premium quality, longer render time |
While $0.10–$0.30/second may seem high, the transcript argues it’s justified: “You can test so many more ads by using AI than by hiring real creators.” For agencies, this cost is negligible compared to production fees.
Current Limitations and Workarounds
No system is perfect. Here are the known limitations:
1. Character Consistency in Video
Issue: OpenAI doesn’t allow face images as references for Sora 2, so human characters may not be consistent across clips.
Workaround: Focus on product consistency (e.g., branded bottles, packaging), which works well—as shown in the hot sauce demo.
2. Token Costs
Issue: Video generation is expensive.
Mitigation: Use Sora 2 for testing, Sora 2 Pro only for final high-stakes campaigns.
3. First-Attempt Imperfection
Issue: AI video models often miss timing or delivery.
Solution: Leverage the agent’s iterative editing tools—trim, replace, remix until perfect.
How the Agent Was Trained: Reverse-Engineering a $2.5M Ad System
The creators didn’t invent a new ad strategy—they reverse-engineered a proven one.
Source of Training Data
A real e-commerce marketer shared a free YouTube video$2.5 million using Sora 2 ads. His process included:
- A 5-step ad creation workflow
- Foundational documents (offer briefs, avatar sheets, product specs)
- Claude-assisted market research before video prompting
Training Methodology
- Downloaded the creator’s public documents
- Copied the video transcript
- Pasted it into ChatGPT with the prompt: “Extract this as a Mermaid graph”
- Imported the graph into Excalidraw for visual workflow mapping
- Built three specialized agents that follow the exact process via handoffs (not parallel collaboration)
Why handoffs? Because “each step of this process requires tight feedback from the user.”
Agent Architecture: Why Three Agents (Not One)?
The system uses a modular agent design for clarity and control:
| Agent | Role | Key Tools |
|---|---|---|
| Strategy Agent | Market research, avatar creation, competitor analysis | Web search, Read Foundational Document tool, document generation |
| Brand Agent | Storyboard and scriptwriting | Template filling, tone adjustment, length control |
| UGC Agent | Video/image generation, editing, final output | Sora 2 API, Gemini 3 Pro, video stitching, subtitle generation |
This separation ensures each phase gets deep focus—and allows users to approve or refine at each stage.
Future Roadmap: What’s Coming Next
The team plans to enhance the agent with:
- ✅ Advanced Subtitles – TikTok-style captions with animations
- ✅ B-Roll Overlays – Add supplementary footage with voiceovers
- ✅ Improved Character Consistency – Once OpenAI allows face references
- ✅ Niche Customization Tools – For builders to fork the agent
How to Customize or Resell This Agent
The entire system is built on the open-source Agency Swarm framework. This means:
- ✅ You can fork the code and adapt it to your niche (e.g., real estate, SaaS, fitness)
- ✅ You can resell it to clients as a white-labeled service
- ✅ All links and resources are freely available (check video description)
For builders, the creator promises a follow-up tutorial: “step by step from scratch how you can build and productize a vertical AI agent for any use case.”
Pro Tips for Best Results
Maximize your output with these expert recommendations:
- Provide rich context: Don’t just say “make an ad for my moisturizer.” Specify audience, pain points, brand voice, and business role.
- Always attach your website: Lets the AI research your actual messaging and positioning.
- Use transcription mode for onboarding questions—it’s faster and more natural.
- Iterate early: Review storyboards before video generation to save token costs.
- Choose Sora 2 Pro for big brands, Sora 2 for testing and smaller campaigns.
Tools and Resources Mentioned
Here’s a complete list of platforms and models referenced:
| Tool | Purpose | Link/Notes |
|---|---|---|
| Agency AI | Deployment platform for AI agencies | Free tier available; hosts the UGC Ad Factory |
| Agency Swarm | Open-source framework for multi-agent systems | GitHub repo linked in video description |
| Sora 2 | AI video generation (standard) | $0.10/second via OpenAI API |
| Sora 2 Pro | AI video generation (premium) | $0.30/second; better for high-stakes campaigns |
| Gemini 3 Pro (Nano Banana) | AI image generation | Used for product mockups; “insane” quality |
| OpenAI API | Access to Sora 2 and language models | platform.openai.com |
| Google AI Studio | Access to Gemini models | Get API key in bottom-left corner |
| Excalidraw | Visual workflow mapping | Used to diagram the ad creation process |
Final Thoughts: The Future of Creative Ops Is Autonomous AI
“Built This UGC” isn’t just another AI tool—it’s a paradigm shift in creative operations. By embedding a proven $2.5M ad process into an autonomous, iterative, and user-friendly AI agent, it eliminates the need for:
- Freelance scriptwriters
- Video production teams
- Manual market research
- Ops hires for ad scaling
And with a risk-free 7-day trial and open-source foundation, there’s never been a better time to replace repetitive creative work with intelligent AI employees.
Ready to Launch Your AI Ad Agency?
All links to deploy, customize, or resell the UGC Ad Factory agent are available in the original video’s description. Whether you’re an e-commerce brand, agency owner, or AI builder, this system offers a scalable, cost-effective path to high-performing UGC ads—without hiring a single new employee.
As the creator says: “Thank you. And don’t forget to subscribe.” But more importantly—go build your AI workforce today.
