The advertising industry has a pattern problem. Not a talent problem, not a technology problem. But a pattern problem. Creative teams build ads one at a time. The flaws accumulate across portfolios. Nobody notices until a machine does.

A new AI-driven system developed by Omnicom Advertising subjects campaigns to two forms of scrutiny before they ever reach an audience. The first uses Google’s ABCD framework to evaluate the structural mechanics of video advertising: attention, branding, emotional connection, and calls to action.

The second, a proprietary layer called BraveBot, focuses on something harder to quantify: whether the work is genuinely distinctive or merely optimized to look like everything else already competing for attention.

When tested on a portfolio of assets for du, the system surfaced patterns human reviewers had missed entirely. Seven out of ten videos delayed brand visibility beyond the crucial opening seconds, and several weakened or buried their calls to action.

Individually, none of these issues appeared catastrophic, but collectively, they revealed how easily modern advertising converges toward the same pacing, structure, and creative logic without anyone noticing in real time. After revisions, performance scores jumped from the forties into the seventies.

Those numbers point to a deeper architectural question: whether combining two kinds of intelligence creates something genuinely new or simply a more defensible iteration of the same work.

THE PATTERN PROBLEM

For all the sophistication of modern creative development, one of its most persistent blind spots has been surprisingly simple: the inability to see across a body of work rather than through it. Human creative directors, however experienced, evaluate ads individually and in isolation. Patterns across a portfolio go unnoticed, not because of negligence but because the process was never designed to catch them.

Noah Khan, Chief Innovation Officer at Omnicom Advertising, is direct about what separates this system from the AI tools that came before it. Most systems, he argues, are built to judge rather than guide.

“Most AI tools in advertising give you a binary score: work is either good or it’s bad. They’re reactive. What makes ABCD fundamentally different is that it’s prescriptive and real-time.”

Rather than delivering a verdict, the framework decodes exactly why engagement drops at any given moment, whether the attention hook is weak, the brand is unclear, the emotional connection is missing, or the call to action fails to land.

Layered on top of that structural evaluation is BraveBot, a proprietary system trained on the agency’s own work, which asks a question that performance data alone does not. “Is this work actually distinctive, or just optimized into invisibility?”

Where most AI solutions present performance and differentiation as a trade-off, Khan argues that this system rejects that choice entirely, flagging work that scores well on conventional metrics but risks disappearing in a crowded market. The goal, as he frames it, is work that is both effective and culturally true. “That combination,” Khan says, “isn’t something you can get elsewhere.”

When tested on du’s portfolio, the findings were specific enough to reframe the conversation about what human review actually misses and why. Khan is careful to reject the obvious explanations when asked whether the gaps reflect a skills problem or a bandwidth problem.

“It’s a pattern-gap problem, and it’s actually revealing about how creative development works.” Human creative directors, he argues, are genuinely exceptional at evaluating individual pieces of work. The problem is not ability but architecture: structural patterns across a portfolio are nearly impossible to catch manually in real time.

The du case made that visible. Seven out of ten videos buried the brand mention after the first five seconds, and three had calls to action that were too weak to register. Creative teams had flagged the issue in one or two videos each, but across the portfolio, it had remained invisible.

After the system flagged them, the fixes were straightforward: brand mentions were moved into the first five seconds, calls to action were made stronger and more prominent, and emotional moments were front-loaded. Videos that had scored between 44 and 48 percent moved into the 72 to 80 percent range. “That’s not replacing creative directors,” Khan says. “That’s giving them pattern intelligence in real time, something human review simply can’t deliver at scale. It’s bandwidth plus insight working together.”

Ibrahim Al Mayahi, Head of Commercial Brand and Marketing Communications at du, frames the value of the process less in terms of what the technology caught and more in terms of when. Operating in one of the region’s most contested telecoms markets, the most meaningful shift for du was to move creative evaluation earlier, to a stage where the work could still be meaningfully changed.

The insights gathered, he explains, helped identify opportunities to improve clarity, branding, and audience engagement across assets before budgets were locked and media spend was committed.

For Al Mayahi, the framing of optimization as a threat to creative instinct misreads what the process actually does. “Optimization is not about replacing creative instinct, but complementing it with data-led insight that helps maximize effectiveness before significant media investment is committed.” The ambition he describes is less a single intervention than a structural one: a faster optimization cycle that, over time, builds a culture of continuous improvement, one that supports stronger campaign performance not just at launch but beyond.

OPTIMIZED INTO INVISIBILITY

Training an AI to recognize courage in creative work is, on its face, a strange proposition. Khan is straightforward about how BraveBot was built to do it. Over decades of work, the agency’s collective has produced, specifically the campaigns it holds up as best in class, the work that actually broke through. “We have measures, frameworks, and taste built in,” he says, which is perhaps the most compressed description of what distinguishes this from a conventional optimization tool.

The question of brave versus merely strange, he argues, is less of a binary than it appears. “He cites a recent Skittles Super Bowl spot featuring Elijah Wood, produced by TBWA\RAAD, as an example.”
. “It’s genuinely weird and genuinely wonderful. It’s smart, funny, and brave, precisely because it risks looking strange.”

When BraveBot and ABCD produce conflicting signals, the resolution is not algorithmic. It comes down to the creative director, whose judgment remains the final authority regardless of what either system recommends.

“These tools exist to support our creative talent, not replace their instincts. We’re augmenting their capabilities, giving them real-time intelligence they didn’t have before. But at the end of every interaction, the creative director has the final say. That’s non-negotiable.”

It is a tension that the framework’s own architects have clearly considered. Aishi Lahiri, Director of Advertising Solutions at Google MENA, is candid about both the scale of the data underpinning ABCD and the limits of what that data should be asked to do. The framework draws on millions of data points from over 11,000 top-performing YouTube ads across 11 verticals and 11 countries and was developed in collaboration with Kantar. It is optimized for the proven structural factors that sustain viewership and drive platform engagement.

But Lahiri is careful to draw a line between structural effectiveness and creative prediction.

“Crucially, we are not using AI to predict creativity,” she says, acknowledging that over-reliance on historical data carries its own risk: the gradual production of what she calls a “sea of sameness.”
The framework handles what she describes as the science of attention, the baseline mechanics that give a piece of work a fighting chance of being watched.

Everything beyond that, the creative ambition, the willingness to defy established norms, remains the province of human judgment. “Human ingenuity remains indispensable in exploring new creative landscapes.” The real impact, as she frames it, comes not from choosing between the two but from the point where AI-driven optimization and bold creative vision meet.

That boundary between tool and talent is something Al Mayahi thinks about from a different vantage point. Speed is the benefit most commonly claimed for AI-enabled workflows, and also the one that makes creative teams most uncomfortable. His response to the tension is to reframe what speed actually means in this context.

The role of AI, he argues, is not to accelerate creativity at the expense of quality but to remove the friction that surrounds it. Strong ideas still require strategic thinking, cultural understanding, and craft, and none of that changes.

What changes is the ability to test, learn, and refine more dynamically before work goes live, giving ambitious ideas, as he puts it, “a stronger chance of succeeding by enabling teams to optimize with greater confidence and precision.”

The future he describes is not one in which speed and creativity are traded off against each other, but one in which the workflow is built so that neither has to be sacrificed. The promise, as he frames it, is not about efficiency but about where creative energy is directed: away from repetitive, manual processes and towards stronger ideas.

WHY INSTINCT STILL MATTERS

It is the question the entire conversation eventually arrives at. If a system can identify a weak hook, push the brand earlier, and prescribe a stronger call to action, the implied threat to the creative director’s role is obvious. Khan’s answer is immediate and unhedged. “The idea itself. That’s what’s left.”

The distinction he draws is between optimization and invention. AI can flag what works based on past performance data, timing, branding mechanics, emotional triggers, and call-to-action construction. What it cannot do is decide what kind of idea to make in the first place. That, he argues, is conception, vision, and taste, none of which is legible to a system trained on what has already existed.

“The most impactful work often looks wrong by conventional metrics because it creates something genuinely new, rooted in human judgment, experience, and the willingness to be brave enough to pursue an idea even when the data looks skeptical at first.”

The formulation he offers is clean enough to double as a design principle: AI optimizes what exists, creativity invents what should exist next, and the two are not interchangeable. The system exists to make sure the best ideas get the execution they deserve, and that weaker ones do not consume client budget, as we found out the hard way.

Lahiri approaches the same question from a different angle, and her answer is less about creative philosophy than operational reality. Google’s ABCD framework has existed long enough to have become a standard reference point for creative evaluation on YouTube. What has not existed, until now, is a production workflow built around it. “This collaboration represents a pivot in how the ABCD framework is being operationalized globally,” she says, acknowledging that knowing a framework and building a live system around it are meaningfully different things.

The division of labor she describes is precise. Google provides the foundational, scalable technology. Omnicom has engineered a bespoke, proprietary workflow that pulls ABCD insights directly into the pre-production lifecycle, rather than applying them after the fact when the work is already done and the budget already spent.

The result is the ability to intervene at the point where it matters most. “Partnering with Omnicom allows us to operate at incredible speed and intercept the process right at the ideation stage, maximizing YouTube creative effectiveness across the market from a campaign’s inception.”

What the collaboration quietly implies about how Google’s broader client base has been engaging with the framework is largely unspoken. However, the fact that it took a regional agency partner to push it into production speaks for itself.