• by Maarten Albarda , Featured Contributor, 7 hours ago

One of my mentors died. You may not have heard of Albert Winninghoff, but after co-founding Noordervliet & Winninghoff in Amsterdam in the 1970s and building it into a premier Dutch agency, Winninghoff sold it to Leo Burnett in the 1980’s. Following the acquisition, he joined Leo Burnett’s global board in Chicago as Vice Chairman.

In that role, Winninghoff helped transform Leo Burnett from an iconic US-centric agency into a global network. He made me media director in his Amsterdam business when I did not think I was ready for it (I was in my late 20s). He then pushed me to take on more international work. He was instrumental in bringing my career out of Amsterdam into the world. He pushed teams to both trust themselves and always do better. He was the embodiment of Leo Burnett’s “Reach for the stars” ambition.

It was in this context that I was thinking about the emerging “sea of sameness” danger of the AI race in our industry. It does not matter if you or your agency uses ChatGPT or Claude; algorithmic convergence guarantees strategic and creative mediocrity for several reasons.

Large language models predict the most probable next token based on historic training data. When agencies and marketers use generative tools to write briefs, brainstorm creative concepts, or build buyer personas, the model outputs the statistical center of our industry’s consensus. It delivers mathematical averageness.

When every agency lets machine learning handle targeting and bidding, every brand bids on the same users at the same times using the same automated adjustments. Hold on, I hear you say, our target audience is highly unique and differentiated from competitors. I hate to disappoint you -- but it isn’t. And even if your audience has different nuances, LLMs will flatten and equalize them really quickly. Major agency platforms (WPP Open, Publicis Marcel, Omnicom Omni) share underlying LLM frameworks, APIs, and data integrations -- and when everyone uses a similar approach you can guess the outcome.

Algorithms are optimized based on historical data and predictable patterns. Breakthrough marketing frequently relies on doing what the data says shouldn't work: taking an asymmetric risk, adopting a polarizing brand voice, or zigging when the tools advise zagging. Brands that strictly follow AI recommendations eliminate the possibility of being an outlier.

So what should you do? You could consider looking for a truly differentiated boutique agency, that relies on a stable of remarkable talent, and forbid them to use AI. I’m sure you can quickly figure out the problems with this approach: Operationally, a complete ban on AI is inefficient. You end up paying $300 to $500-a-hour senior talent to execute routine data formatting, tag checks, and image resizing.

A better counterstrategy could be to ban AI from the “upstream phase” (strategic planning, brand positioning, master creative concepts) but deploy it fully in downstream (asset variations, localizations, data piping).

If you do, make sure you take note of the recent World Federation of Advertisers “Global AI Adoption and Agency Remuneration Benchmarks” study. It reveals that 96% of major global brands now deploy generative or agentic AI across their marketing operations. However, many advertisers pay their agencies both a new AI platform access fee as well as full-time equivalent (FTE) staffing headcount. Few agencies provide verified evidence showing that software contribution reduced billable human hours. Even fewer can demonstrate that AI “enhanced” work is, at minimum, equally good or better than what you had before.