Shai Luft, Co-Founder and COO, Bench Media
AI-driven advertising has moved beyond experimentation and is now embedded in how media is planned, bought and optimised.
Across major platforms, automation is no longer framed as an optional enhancement but increasingly as the default, which means the conversation for agencies and brands has shifted from whether AI should be used to how much control, visibility and understanding they are prepared to give up in exchange for efficiency.
From an agency perspective, this tension is becoming harder to ignore.
On one hand, AI-led products are delivering genuine performance improvements, with faster learning cycles, broader reach and lower costs than many manual approaches could achieve.
On the other hand, those same systems are steadily reducing the ability for agencies and brands to influence how their advertising behaves in market, including how messaging appears, where it appears and the context in which it is experienced.
The struggle for control
Much of the debate has focused on control, and rightly so.
As AI-driven campaign structures become more prevalent, agencies are finding it harder to ensure creative remains on brand, placements align with brand values and messaging feels intentional rather than incidental.
Creative combinations are assembled dynamically, placements are optimised at scale and budget decisions are increasingly made within systems that offer limited transparency.
Agencies remain accountable for outcomes, yet their ability to shape or interrogate those outcomes continues to narrow.
This challenge is particularly evident on the largest global platforms, including Google and Meta, where the shift towards AI-led execution has accelerated rapidly and, in many cases, become the default rather than the exception.
While this approach has unlocked efficiency gains, it has also made it more difficult for agencies and brands to control messaging, placements and experimentation in a deliberate way.
Campaigns can behave differently from how they were originally intended, often with limited visibility into which automated decisions are driving performance.
The issue is not only strategic but operational. AI features are frequently enabled by default or applied broadly once a campaign type is selected, which makes it easy for automation to influence outcomes unintentionally.
In practice, this means campaigns can go live with AI-driven elements shaping delivery without that behaviour being explicitly planned or approved.
Intent vs. blind automation
From an agency point of view, that introduces risk, because intent matters. Advertising should behave the way it was designed to behave. In theory, AI can be extremely powerful when used in a controlled and deliberate way.
Testing individual AI capabilities in isolation allows agencies to understand their impact and build meaningful insight over time.
That kind of structured experimentation is how performance improves sustainably and how learning compounds across campaigns and years.
In reality, however, the way AI is being rolled out is making this increasingly difficult, with controls that are easy to miss or automation that applies across entire campaigns, leaving little room for incremental testing or selective oversight.
The erosion of learning
Beneath the issues of control sits a deeper concern that goes to the heart of marketing as a discipline: learning.
The most valuable asset marketers develop over time is not efficiency alone, but understanding. Understanding what drives consumer behaviour, which messages resonate and why, and which triggers matter at different moments in the journey.
This accumulated learning is what drives marketing progress. When AI-driven systems obscure cause and effect, that learning loop weakens. Campaigns may perform well, but if it is unclear what actually drove the outcome, performance improves without understanding, and long-term progress becomes harder to sustain.
AI has the potential to accelerate marketing intelligence by helping teams surface patterns faster and navigate complexity more effectively. That potential is only realised when automation enhances human judgement rather than replacing it.
When AI limits visibility and explanation, it risks delivering outcomes without insight, leaving agencies and brands disconnected from the signals they need to improve.
Finding the balance
This is not an argument against AI advertising. The technology is powerful and the benefits are real.
It is, however, an argument for balance. Brands and agencies should not have to choose between results and understanding, between efficiency and learning, or between automation and intent.
Marketing only moves forward when learning compounds over time. If AI-driven advertising erodes that foundation, the industry risks becoming faster and cheaper, but also less thoughtful, less differentiated and ultimately less effective in the long run.
That is a trade-off worth questioning before it quietly becomes the default.
Main image: Shai Luft, Co-Founder and COO, Bench Media.