Launch This Free UGC Ad Agent in 60 Seconds

Key Concepts

• AI Employees/Agents: Autonomous AI entities designed to perform specific tasks, replacing human hires for reporting and ad creation.
• UGC Ads (User-Generated Content Ads): Advertisements that mimic the style and authenticity of content created by users.
• Agency Swarm: An open-source framework for building and deploying AI agents.
• Agency AI Platform: A platform for deploying and managing AI agents, including a marketplace.
• Sora 2 / Sora 2 Pro: AI models for video generation.
• Gemini 3 Pro / Nano Banana: AI models for image generation.
• Persistent File Storage: Enables AI agents to save and reuse files across conversations.
• Custom GPTs: Whitelabeled custom web applications built using AI agents.
• Foundational Market Research Documents: AI-generated documents providing insights into product positioning, target audience, and competitors.
• Storyboard Generation: AI-created visual outlines and scripts for video ads.
• UGC Ad Creator Agent: The final AI agent responsible for generating video prompts and the videos themselves.
• Iterative Refinement: The ability to provide feedback and adjust AI-generated results.
• Character Consistency: A current limitation in AI video generation, particularly with faces.
• Token Cost: The expense associated with using AI models, especially for video generation.
• Vertical Agent: An AI agent trained on a specific business process or niche.

AI-Powered UGC Ad Factory

This video introduces an AI-driven system designed to autonomously generate User-Generated Content (UGC) ads, aiming to eliminate the need for additional human hires in operations and reporting. The system offers a 7-day free proof of concept, with no payment required if the service doesn't prove its value.

Core Functionality and Process

The entire ad creation process, from market research to scripting and video generation, is handled by AI agents autonomously. The workflow is as follows:

1. Product Information Input: Users begin by answering a few questions about their product.
2. Market Research Generation: A "strategy agent" generates four foundational market research documents. These documents are crucial for understanding the product's market position and target audience.
3. Storyboard and Script Creation: The system then transitions to another agent that generates a storyboard with accompanying scripts.
4. UGC Ad Creation: Upon confirmation of the storyboard, a final "UGC ad creator agent" takes over. This agent generates prompts for video generation, creates the videos themselves using AI models like Sora 2, stitches them together, and adds subtitles, resulting in a ready-to-use ad.

An example provided is an ad for "Herbalux," an organic, fragrance-free aloe moisturizer, highlighting how a damaged skin barrier reacts to irritants and positioning Herbalux as a solution.

Differentiators from Existing UGC Systems

This AI system distinguishes itself from other UGC automation tools in several key ways:

• Real-World Training Data: It was trained on the actual processes of an e-commerce agency that generated over $2.5 million with Sora 2 ads.
• Iterative Refinement: Unlike simple automation, this system allows for continuous refinement of results based on user feedback. Users can request multiple product images, ask for specific segments to be replaced or remixed, and generally iterate with the agent to achieve desired outcomes.
• Organized Asset Management: Videos are organized into folders, mimicking the functionality of a professional video editor.

Deployment and Setup

The AI agent is built on the open-source framework "Agency Swarm" and is accessible through the "Agency AI" platform.

Deployment Steps:

1. Access Agency AI Marketplace: Navigate to the Agency AI platform and its marketplace.
2. Select UGC Ad Factory Agent: Choose the "UGC Ad Factory" agent.
3. Model Selection:
   • Video Generation: Sora 2 is recommended for decent results; Sora 2 Pro is suggested for larger brands.
   • Image Generation: Gemini 3 Pro image preview (Nano Banana model) is recommended.
4. Upload Logo: Upload the brand's logo.
5. API Key Integration:
   • OpenAI API Key: Obtain from platform.openai.com.
   • Gemini API Key: Obtain from Google AI Studio.
6. Save and Deploy: Paste the API keys and save. Deployment takes approximately one minute.
7. Enable Persistent File Storage: This feature allows the agent to save and reuse files across conversations.
8. Integrate into Custom GPT: Deploy the agency within a custom GPT, creating a whitelabeled web application.

Practical Application: Hot Sauce Ad Example

The video demonstrates the process using a hypothetical hot sauce product called "Spicy Prompt" for an "EI agency."

Detailed Workflow with "Spicy Prompt":

1. Onboarding Questions: The agent asks about the product's role, target audience, brand identity, and website.
   • Product Role: Gift for clients after engagements.
   • Target Audience: Marketing agency owners earning $500K+.
   • Brand Identity: Playful but sharp.
   • Website: Provided for research.
2. Foundational Document Generation:
   • The agent analyzes the provided website and company details.
   • It performs web searches to understand competitors.
   • It generates a comprehensive "product document" containing brand details for future ad creation.
   • It populates "offer brief" and "avatar sheet" templates based on the e-commerce agency's templates.
   • Four foundational documents are created and organized in agent storage. One document details customer beliefs based on market research.
3. Storyboard and Script Iteration:
   • The "strategy agent" transfers to the "brand agent" for storyboard generation.
   • An initial 80-second storyboard with scene descriptions and scripts is created.
   • User Feedback: The user requests the storyboard to be funnier and shortened to 30 seconds. The agent successfully refines it. The speaker notes that GPT 5.1 is funnier than GPT 5.
4. Video Generation and Refinement:
   • The "brand agent" transfers to the "UGC agent."
   • The agent lists brand assets and generates product images (two variants) for reference, incorporating brand colors and the "AI infused" text.
   • Video Creation: The agent uses the "generate video" tool with Sora 2, creating prompts based on the refined storyboard. The prompts are described as comprehensive, saving hours of manual work.
   • Three videos are generated in parallel.
   • Example Ad Copy: "Most agencies send a polite thank you email. We send a bottle of hot sauce so aggressive it should come with HR training. We call it spicy prompt."
   • User Feedback: The second video clip is identified as having an insufficient length and a cut-off script. The user instructs the agent to fix it and add the logo.
   • Logo Integration: The logo is successfully added to the hot sauce bottle in the revised video.
   • Final Edits: The user requests combining the videos and adding subtitles.
   • Further Refinement: A minor issue at the end of the third video is addressed by asking the agent to remove one second from the last video, demonstrating the agent's video editing capabilities.

Limitations and Future Development

• Character Consistency: Currently, OpenAI does not allow images with faces as references for the Sora 2 model, impacting character consistency. However, product consistency is achievable.
• Token Cost: Video generation can incur significant token costs, though it's presented as a cost-effective alternative to hiring human creators. The cost for the hot sauce video was approximately $9, with Sora 2 accounting for over 90% of this.
  • Sora 2 API: $0.10 per second.
  • Sora 2 Pro API: $0.30 per second.
• Future Enhancements:
  • Improved subtitle generation (e.g., TikTok-style).
  • Ability to add video overlays with B-rolls and voiceovers.

Training Methodology

The AI agent was trained on a real business process derived from a YouTube video by an e-commerce agency owner who generated $2.5 million with Sora 2 ads.

Training Process:

1. Source Material: The e-commerce agency owner's video detailing a five-step process and sharing foundational documents used for pre-prompt research.
2. Document Download: The foundational documents shared by the agency owner were downloaded.
3. Agent Training: The AI agents were trained to follow this process, accessing the foundational documents via a "read foundational document" tool.
4. Process Extraction: The transcript of the e-commerce agency owner's video was copied and pasted into ChatGPT, with instructions to extract it as a Mermaid graph.
5. Visualization and Training: The extracted process was then inserted into Excali draw for visualization and subsequently used to train the AI system.
6. Agent Structure: The system was designed with three agents, using "handoffs" for communication. This means agents replace themselves in the conversation history rather than collaborating directly, as each step requires tight user feedback.

Conclusion and Call to Action

The system offers a powerful and efficient way to generate high-quality UGC ads autonomously. The ability to iterate and refine results, combined with the underlying framework's open-source nature, makes it a valuable tool for businesses. The video encourages viewers to try the agent for free, explore the Agency AI platform, and provides links for further resources, including a video on building custom AI agents.