In the rapidly evolving landscape of digital advertising, small-to-medium brands face a daunting dilemma: how to keep pace with the massive creative volume demanded by Meta’s algorithmic updates without burning out a design team or shattering the marketing budget. With the implementation of the "Andromeda" update—which effectively consolidated hundreds of slight ad variations into single creative units—the "spray and pray" approach is officially dead.
Today, success requires high-quality, distinct creative assets that resonate deeply with target audiences. For many, the answer lies in Artificial Intelligence. However, as Fraser Cottrell, CEO of the direct-to-consumer agency Fraggell, suggests, most marketers are approaching the technology with two critical misconceptions: that AI is a "lazy" shortcut, and that it inherently produces low-quality visuals.
In reality, AI is a precision instrument. When utilized correctly, it levels the playing field, allowing smaller brands to produce studio-quality assets for mere cents on the dollar. This article outlines the three-step framework for mastering AI-driven creative at scale.
1. Building the Foundation: Deep Research and Knowledge Bases
Before generating a single image or video, marketers must bridge the gap between generic AI outputs and brand-specific brilliance. The foundational step is the creation of a "Brand Knowledge Base."
The "Deep Research" Methodology
Fraser Cottrell emphasizes that the quality of AI output is entirely dependent on the context provided. When onboarding a new client, Fraggell initiates a "deep research" phase. Using Large Language Models (LLMs) like Google’s Gemini, the team prompts the AI to act as a market researcher, scanning the internet for comprehensive brand profiles.
The objective is to move beyond superficial demographic data. The research must uncover:
- The "Why Not" Factor: Why do potential customers encounter the product but choose not to purchase?
- Customer Sentiment: What are the recurring complaints on Reddit, niche forums, and review platforms?
- Pain Points: Where are the friction points that prevent a conversion?
To automate this, marketers can use voice-to-text tools like Whisper Flow to dictate complex prompts to AI models, ensuring the research covers geographical concentrations, competitor objections, and buyer motivations.
Verifying and Supplementing Data
AI research is an excellent starting point, but it is not infallible. Once a document is generated, it must be verified. A highly effective technique is to feed the generated research back into a secondary model (such as Claude) and instruct it to act as a devil’s advocate. By asking the AI to "interview" you about the document, you can systematically flag inaccuracies or gaps in the research.
Finally, you must inject "human-only" intelligence. The internet doesn’t know your proprietary customer data, the nuances of your product’s internal testing, or the specific language your best customers use during support calls. Integrating this internal knowledge with AI-driven external research creates a unique, proprietary knowledge base that no competitor can replicate.
2. Training a Dedicated "Claude Project"
Once your data is validated, the next step is centralizing it. Claude’s "Projects" feature offers a dedicated workspace with persistent memory—a private ecosystem for your brand’s creative identity.
What to Feed the AI
A Project is only as strong as the data it consumes. To build a high-performing creative engine, load the following assets into your Claude Project:
- The Deep Research Document: The corrected, human-verified version from Step 1.
- Voice-of-Customer (VoC) Data: Exported reviews and testimonials. This language is critical, as it mimics the authentic vernacular of your buyers.
- Internal Brand Guidelines: A manifesto detailing the brand’s voice, visual standards, and, crucially, a definition of what constitutes a "good" ad.
- Performance Data with Visual Context: Don’t just rely on spreadsheets. Use tools like Poppy or Gemini’s vision capabilities to analyze your top-performing ads from the previous quarter. By letting the AI "watch" your successful videos and correlate them with performance metrics, you teach the model the specific visual language—pacing, hook style, and on-screen action—that drives your specific audience to act.
This setup ensures that every future prompt—whether for a headline, an image, or a video script—is filtered through the lens of your brand’s history and performance data.
3. The Hybrid Creative Workflow: Scaling Production
With the brand context established, you are ready to transition from planning to execution. A "hybrid" approach—using AI for ideation and structure, while keeping human oversight for final assembly—remains the gold standard for quality control.
Static Image Generation
For static ads, the most efficient workflow involves using AI to generate the visual and keeping the text layer separate. By using tools like Nano Banana 2 Pro (within Gemini) or via API integrations, you can generate multiple variations of a product shot simultaneously.
- The Prompting Strategy: Use the Claude Project to brainstorm headlines based on the VoC data you’ve uploaded. Once you have a selection, use the project to draft specific prompts for image generation.
- Iterative Refinement: If an image doesn’t hit the mark, provide specific feedback to the AI. Tell it what you like and what you dislike. Because the Claude Project retains conversation history, the AI learns your preferences progressively, becoming more "on-brand" with every iteration.
Video Scripting and Ideation
While current AI video tools are improving, the gold standard for high-performance video remains human-shot or professional UGC (User Generated Content). However, AI is a force multiplier for the scripting process.
By describing your desired scenario—who is in the video, the setting, and the target length—to your trained Claude Project, you can receive a professional-grade, timestamped script in seconds. Even if the AI’s output isn’t perfect, it typically covers 70% of the heavy lifting. A human creative can then refine the tone, tweak the hook, and polish the emotional beats, resulting in a finished product in a fraction of the time it would take to start from a blank page.
Implications: The New Competitive Landscape
The shift toward AI-assisted creative is not merely a trend; it is a structural change in how marketing teams operate.
Leveling the Playing Field
Historically, the barrier to entry for high-quality ad creative was steep. It required professional photography, studio rentals, and expensive post-production. Today, those costs are negligible. A small, scrappy brand with a well-trained AI knowledge base can now compete visually with enterprise-level companies.
The Death of "Quantity Over Quality"
Meta’s move away from high-variation, low-quality testing means that every ad you launch must be a potential winner. The "AI-first" methodology allows for the rapid testing of distinct creative concepts rather than minor tweaks. By understanding the "why" behind their customers’ pain points—as identified in the Deep Research phase—advertisers can focus their efforts on creating assets that genuinely solve problems, rather than cluttering the feed with noise.
The Role of the Human Marketer
The most significant implication is the evolution of the marketer’s role. The creative process is no longer about manual labor; it is about curation, strategy, and verification. The marketer becomes an editor-in-chief, responsible for training the AI, vetting the research, and ensuring that the final output resonates with the human element of the brand.
Conclusion
The era of burnout-inducing creative production is ending. By systematically building a brand knowledge base, training persistent AI projects, and adopting a hybrid creative workflow, modern advertisers can generate high-performance assets at the scale required by today’s algorithms.
As Fraser Cottrell notes, getting AI to produce what you actually want requires significant effort—but that effort is now directed at strategy rather than execution. For the brand that invests the time to train its AI properly, the result is not just a faster workflow, but a significant competitive advantage in an increasingly crowded digital marketplace.
