In the modern digital landscape, the demands of advertising algorithms—specifically Meta’s recent "Andromeda" update—have fundamentally altered the requirements for brand survival. Where marketers once relied on launching hundreds of minor ad variations, platforms now favor unified, high-quality creative assets. For small brands and lean design teams, this shift presents a daunting challenge: how to maintain consistent, high-performing output without the crushing overhead of traditional production.
According to Fraser Cottrell, CEO of the direct-to-consumer agency Fraggell, the solution is not to work harder, but to work smarter by integrating generative AI into the creative workflow. By moving past common misconceptions—namely that AI is "lazy" or inherently low-quality—marketers can leverage these tools to level the playing field, generating professional-grade creative at a fraction of the traditional cost.
The Shift in Ad Strategy: Quality Over Quantity
For years, the standard playbook for Meta advertising involved aggressive A/B testing, where advertisers flooded the system with hundreds of nearly identical variations. Meta’s latest updates have rendered this obsolete; the algorithm now groups those variations into a single creative entity. Consequently, the burden has shifted to the quality of the initial asset.
Brands that once spent thousands on studio photography or freelance designers can now utilize generative AI to produce high-fidelity imagery for mere cents. However, this transition requires a shift in mindset. AI is not a "magic button" that creates successful ads in isolation; it is a sophisticated tool that requires rigorous training, context, and iterative human oversight.
Step 1: Building a Foundational Knowledge Base
Before generating a single pixel, marketers must establish a "Brand Knowledge Base." As Cottrell notes, AI is only as effective as the context provided to it.
The Deep Research Phase
The process begins with "Deep Research," a function available in advanced LLMs like Google Gemini. Unlike a standard search, a deep research prompt instructs the AI to browse the web, analyze Reddit discussions, identify customer pain points, and synthesize competitor strategies.
To execute this, marketers should use a structured prompt:
- Define the Goal: Instruct the AI to build a comprehensive profile of the brand.
- Target the "Why": Analyze not only why customers purchase but also why potential customers abandon the funnel.
- Surface Objections: Extract common complaints and geographical concentrations of the audience.
Once the AI generates this document, verification is critical. Cottrell suggests a "verification loop" where the document is fed into Claude with instructions to act as a skeptical editor, asking the user questions to confirm or correct key facts. This ensures the foundational data is accurate before it informs future creative decisions.
Integrating Proprietary Data
Public data from the internet only scratches the surface. The most effective ad creative stems from proprietary insights—internal sales data, customer service call transcripts, and nuanced product knowledge that AI cannot "scrape." By manually injecting these internal insights into the research document, marketers create a hybrid knowledge base that is both broad (market-wide) and deep (brand-specific).
Step 2: Training a Dedicated Claude Project
With the research document finalized, the next stage involves creating a "Claude Project." Unlike a standard chatbot session, a Project serves as a persistent workspace with its own long-term memory. It acts as a dedicated repository for brand identity.
Curating the Project Library
To optimize the AI’s performance, load the following assets into the project:
- The Verified Research Document: The core summary of market position and customer pain points.
- Voice-of-Customer (VoC) Data: Exported CSVs of customer reviews and testimonials. This language is the most powerful tool for copywriting.
- The Internal Brand Bible: A document outlining the brand’s tone, aesthetic guidelines, and definition of a "good" ad.
- Performance Analytics: Historical ad data from the previous quarter, paired with visual context.
For the visual element, tools like Poppy are instrumental. By analyzing the visual pacing and on-screen action of top-performing videos, marketers can feed those insights into the Claude project. This allows the AI to understand not just the data behind a winner, but the visual language that resonated with the audience.
Step 3: Executing Creative Production
With the knowledge base established, the production phase becomes an iterative, human-led process.
The Hybrid Approach to Image Ads
Cottrell advocates for a "hybrid approach": generating the image via AI, while handling the text manually. This allows for rapid testing of different copy variations against a single, high-quality visual without requiring a complete regeneration of the asset.
When brainstorming, the prompt should be treated as a creative brief. For instance, if targeting marathon runners for a hydration product, specify the persona and the specific objection the ad must overcome. If a reference ad is available, upload it to the AI and ask it to adapt the concept for your brand, ensuring the output remains distinct while benefiting from proven psychological triggers.
The Role of AI in Video Scripting
While current AI video generation tools are still evolving, AI excels at the scripting and ideation phase. By describing the scenario—the characters, the setting, and the desired length—to the trained Claude project, marketers can receive a detailed, timestamped script in seconds.
This script should be treated as a "30% draft." While the AI provides the structure and the hooks, a human copywriter must refine the tone to ensure it maintains the brand’s authentic voice. This workflow reduces the time-to-production for video concepts by hours, if not days, allowing creative teams to focus on high-level strategy rather than staring at a blank page.
Implications: The Democratization of Creative
The move toward AI-integrated creative has profound implications for the advertising industry.
- Leveling the Playing Field: The barrier to entry for high-quality production is collapsing. Small, agile brands can now compete with enterprise budgets by using AI to generate high volumes of diverse creative, testing more concepts than a traditional agency could produce in a month.
- The Shift to Curation: The role of the marketer is evolving from "creator" to "curator." Success now depends on one’s ability to curate data, guide the AI, and exercise human judgment to filter out the generic.
- Algorithm Alignment: As Meta and other platforms refine their algorithms to value creative quality over brute-force variation, this systematic approach to AI training ensures that brands remain aligned with platform best practices.
Conclusion: A New Era of Efficiency
The era of manual, labor-intensive ad production is fading. By viewing AI as a partner in the research, training, and ideation processes, marketers can satisfy the algorithm’s hunger for fresh content without burning out their teams.
As Fraser Cottrell emphasizes, the most successful brands will be those that treat their AI projects as a living repository of brand knowledge. By systematically building this foundation, brands can ensure that their AI-generated output is not only faster and cheaper, but consistently more effective at converting the target audience. The technology is here; the winners will be those who learn how to train it.
