In the rapidly evolving landscape of digital advertising, small-to-midsize brands are facing an existential crisis. Meta’s algorithmic shifts, specifically the transition to the "Andromeda" update, have rendered the old "spray and pray" tactic of running hundreds of slight creative variations obsolete. Today’s platforms demand high-quality, high-volume assets that resonate deeply with audiences, but for many marketing teams, this creates a bottleneck: how do you scale production without burning out your designers or breaking the budget?
According to Fraser Cottrell, CEO of the direct-to-consumer agency Fraggell, the answer lies in a systematic integration of generative AI. Far from being a "lazy" shortcut, leveraging AI to produce ad creative is a sophisticated operational pivot that allows lean teams to compete with industry giants.
The Core Misconceptions of AI in Advertising
To understand the potential of AI, one must first dismantle two pervasive myths that prevent many marketers from taking the leap.
The first misconception is that using AI is a shortcut for the uninspired. Cottrell argues the opposite: achieving high-end results requires immense intentionality. "Getting AI to produce what you actually want during the creative process requires significant effort," he notes. It is a tool for orchestration, not a magic button.
The second, perhaps more damaging, myth is the belief that AI produces low-quality creative. While video generation still has room to evolve, static image generation has reached a tipping point. When provided with the right context and precise instructions, AI models can produce images nearly indistinguishable from professional studio photography. For e-commerce brands, this levels the playing field, turning what was once a multi-thousand-dollar studio expense into a process that costs pennies.
Step 1: Building a Foundational Brand Knowledge Base
Before a single image is generated, the marketer must perform "Deep Research." This foundational phase ensures that the AI is not hallucinating or producing generic output, but is instead operating with a granular understanding of the brand’s unique market position.
The Deep Research Methodology
Cottrell utilizes Large Language Models (LLMs) like Google Gemini—preferred for its superior data-browsing capabilities and speed—to conduct comprehensive audits. The goal is to move beyond surface-level insights. A successful deep research session targets:
- Customer Personas: Who is buying, and more importantly, why?
- The "Anti-Persona": Why do potential customers encounter the brand and choose not to buy?
- Sentiment Analysis: Scraping Reddit and other forums for raw customer feedback, common complaints, and unaddressed pain points.
- Geographic Concentration: Identifying where the customer base is most dense to tailor localized creative.
To streamline this, marketers should use tools like Whisper Flow to dictate research prompts to LLMs, ensuring the AI performs a thorough web browse rather than relying on stale training data.
Verifying and Augmenting the Data
AI output is only as good as its verification. Cottrell recommends a "Claude-check" method: once the deep research document is compiled, paste it into a fresh Claude session and instruct the AI to quiz you, one point at a time, on the accuracy of the findings. This iterative process allows you to correct the AI’s assumptions and, crucially, inject "proprietary knowledge"—the human-led insights, internal data, and nuances of product usage that no web scraper could ever uncover.
Step 2: Training a Dedicated Claude Project
Once the research document is polished, it must be localized into a "Claude Project." This is a persistent workspace that acts as a secure, dedicated memory bank for your brand. Unlike a standard chatbot conversation, a project ensures that every interaction is grounded in the same foundational truth.
What to Feed the AI
The quality of your creative output is directly proportional to the "context" you feed the project. Cottrell suggests populating your project with:
- The Verified Deep Research Document: The bedrock of your brand identity.
- Customer Testimonials: Direct exports from your store or review platforms. This provides the AI with "voice-of-customer" language—the exact phrasing your buyers use.
- The Internal Brand Bible: A document outlining brand values, tone, and the specific definition of what constitutes a "good" ad.
- Performance Data with Visual Context: This is the most critical technical step. By using tools like Poppy, which can "watch" video assets to analyze pacing and visual motifs, you can feed the AI insights on why a previous ad succeeded.
Step 3: Execution and Iterative Creation
With a trained project in place, the creative process shifts from "guessing" to "systematic generation."
The Hybrid Approach to Image Ads
For static images, the most effective workflow is a hybrid: generate the imagery via AI, but handle the text/copy overlay manually. This decoupling allows you to swap headlines or calls-to-action without regenerating the entire background asset.
When prompting, provide the AI with specific creative briefs. If you want a studio shot of your product, upload an actual photograph of the item to Claude, then describe the environment, lighting, and mood. The AI will then generate a technical prompt optimized for image-generation tools like Nano Banana 2 Pro or other high-fidelity engines.
Scripting and Video Ideation
While fully AI-generated video is still maturing, AI is a superpower for scripting and storyboarding. By feeding your Claude project a concept—such as "a 30-second UGC script for a marathon runner using our hydration product"—the AI provides a timestamped, structured framework.
"The AI-written script isn’t a finished product," Cottrell clarifies. "It gets a creative person 30% of the way to the finish line in a fraction of the time." A human copywriter can then step in to inject the final 70%—the nuance, the conversational rhythm, and the emotional resonance that AI currently lacks.
Implications for the Future of Ad Agency Operations
The adoption of this three-step system represents a paradigm shift in how marketing teams operate.
Chronology of the New Workflow
- Research (Days 1-2): AI conducts deep market audits; human teams verify and supplement with proprietary data.
- Contextualization (Day 3): All data is moved into a dedicated Project environment.
- Ideation & Scaling (Ongoing): Marketers use the Project to generate dozens of variations, test them, and feed the results back into the Project memory to refine future output.
The Strategic Shift
The implications of this shift are profound. Agencies and in-house teams are moving away from manual, labor-intensive asset creation toward "AI-facilitated creative direction." The marketer’s role is evolving into that of a curator and strategist, where their primary value is no longer the ability to execute the design, but the ability to provide the AI with the right context to execute it for them.
As Meta’s algorithm continues to prioritize cohesive creative strategies over high-volume testing of weak assets, the brands that win will be those that use AI to maintain high quality at scale. By grounding AI in deep research and brand-specific training, marketers can finally stop fighting the algorithm and start working in lockstep with it. The barrier to entry for high-quality, professional-grade advertising has been dismantled—the only remaining barrier is the willingness of the marketer to learn how to orchestrate the machine.
