Going In 2026, the advertising industry is being seduced by a dangerous metric: Cost Per Asset.
On paper, the math looks irresistible. A traditional photoshoot costs $50,000 and takes three weeks. Midjourney or Runway Gen-3 can generate a similar image or video clip for $0.10 in 30 seconds. CFOs look at these numbers and see a revolution.
But marketing leaders are starting to see something else: a massive hidden cost. It’s called "AI Slop"—the flood of generic, glossy, slightlyuncanny content that consumers are learning to ignore (or worse, mock).
The real efficiency in AI isn't about replacing your creative team to save money. It is about leveraging high-level expertise to pilot these tools.
Here is why the most cost-effective workflow for 2025 isn't "AI Only"—it's "AI + Expert."
1. The New Cost Ratio: Expertise as a Multiplier
In the traditional workflow, you paid for time. You paid a retoucher for 10 hours of work to get one perfect image.
In the amateur AI workflow (the "press button" method), you pay almost nothing, but you get zero brand equity because the output looks fake.
The "Expert AI" workflow changes the ratio. An experienced Creative Director using AI doesn't just "make things faster." They use the tool to skip the mediocre parts of the process so they can spend more time on the exceptional parts.
- Traditional: 100 hours = 1 Campaign
- AI (Novice): 1 hour = 100 Generic Assets (Low Value)
- AI (Expert): 10 hours = 5 High-Impact, Custom Campaigns (High Value)
The Insight: You are not paying the expert for the 30 seconds it takes to generate a prompt. You are paying them for the 20 years of taste that tells them which prompt will resonate, and how to fix the result when the AI gets it wrong.
2. The "Sandwich" Method: Why Human Input is Non-Negotiable
To avoid the "slop" that feels disconnected and cold, successful agencies are adopting a "Human Sandwich" workflow. AI is the meat, but humanity is the bread that holds it together.
The Top Bun: Pre-Production & Strategy
You cannot ask an AI to "make a viral ad." You have to guide it with cinematic precision. This requires a human who understands:
- Camera Angles & Lenses: Knowing the difference between a 35mm anamorphic lens and a fisheye is the difference between a cinematic masterpiece and a distorted mess.
- Concepting: AI is a cliche machine. It defaults to the most common denominator. A human strategist is needed to force the AI away from the obvious and toward the unique.
The Meat: The AI Generation
This is where the efficiency happens. We iterate rapidly. instead of waiting days for a render, we see options in seconds. But even here, the human is active—curating, re-rolling, and tweaking parameters.
The Bottom Bun: Post-Processing & "Grounding"
This is the step that 90% of brands miss, and it is why their content fails. Raw AI output rarely has a "soul." It feels floaty and weightless.
- Compositing: A human editor takes the AI background and composites real product photography into it so the product doesn't look hallucinated.
- Color Grading & Grain: AI video is often too clean. A human colorist adds film grain, authentic light leaks, and color consistency to "ground" the footage in reality.
- Sound Design: AI generates video, but rarely good audio. The emotional connection of an ad comes from the human choice of music, sound effects, and voiceover.
3. Avoiding the "Uncanny Valley" of Brand Trust
Consumers have developed a "sixth sense" for AI. When they see a video that was clearly just copy-pasted from a generator, they subconsciously tag the brand as "cheap" or "lazy."
Real connection requires imperfection. It requires a specific brand voice. When you just "press the button," you get the average of the internet. When you have an expert edit and customize that output, you get something that feels intentional.
The Bottom Line
The tools are free (or close to it). The expertise is the premium.
The most efficient agencies of 2025 won't be the ones with the best software; they will be the ones with the best pilots—creatives who know how to take a robot's hallucination and turn it into a human story.
L. Arriaga
(Note: This "Expert + AI" model doesn't just improve quality; it accelerates learning. We can now run A/B tests at a daily velocity that was previously impossible, rapidly refining our approach.
But the reality is: Platforms are getting smarter. I believe future algorithms will prioritize content where they can detect "heavy human input"—nuanced color grading, specific editing rhythms, and complex transitions. While "AI Slop" will flood the feeds, the algorithms (and the audiences) will increasingly filter for craft. The tools will evolve, but the winners will still be the ones with the human expertise to wield them.)