Why Can’t We Fully Automate Media Yet?

Too many variables: audience shifts, inconsistent metrics, and creative impact. Automation optimizes, but strategy needs human inference. The challenge is balancing tech efficiency with smart decision-making.

Last week, a colleague hit me with a question that seemed incredibly simple on the surface: If we know the drivers of impact and the KPIs that influence them, why can’t we automate 99% of media planning, buying, and optimization? It’s not even AI, just basic automation.

I had to pause. Not because the question wasn’t valid—but because answering it in a way that made sense in hyper-rational terms required real thought.

The Problem: Too Many Moving Parts

Media planning isn’t just plugging numbers into an equation. We make assumptions about how different variables will interact, but we don’t know their true impact until they’re in-market. Here’s why:

  1. Audience & Circumstance Dependence – A campaign’s effectiveness varies based on product, audience perception, competition, and macroeconomic shifts. The same strategy can perform brilliantly in one context and flop in another.
  2. Metrics Aren’t Apples-to-Apples – A “TV impression” is often a 15-30 second forced view. A “display impression”? 50% of pixels on screen for 2 seconds. One might play across the room while you’re making coffee, the other inches from your face. Same term, different realities.
  3. Siloed & Overlapping Data – Each platform provides reach estimates, but since they guard their user data, we can only guess at audience overlap.
  4. Creative Variability – Ads aren’t interchangeable widgets. Performance depends on format, campaign, brand, and countless other factors—making automation harder than it looks.

But What About Optimization?

Once media is live, sure, we automate a lot. Digital platforms and even traditional media now goal-seek toward KPIs, which is why conversion tracking is of paramount importance.

But no external media partner sees all of a company’s media spend. It’s up to the company to merge that data and assess true incremental impact—matching media exposure to business outcomes.

Can’t We Just Build a Model?

Yes… and no.

Marketing changes hearts and minds. It influences behavior in ways that don’t have an instant feedback loop. Since we don’t have brain chips measuring real-time persuasion, we rely on probabilistic models, not deterministic ones.

Did you buy that expensive B2B software because of the TV ad, the email, or the sales call? We can estimate likelihoods, but we’ll never know definitively.

Why Humans, Not Machines?

At the highest level, media strategy is about inference, not just optimization.

  • Humans recognize patterns, extrapolate meaning, and shift perspectives fluidly. Machines struggle with vague, high-level objectives.
  • Teams align toward broad goals, even amid ambiguity. Machines optimize only for the clearly defined desired outcome.

But Here’s the Catch…

Because the system is so complex and imperfect, we often over-rely on human judgment and adaptability. And while that brings flexibility, it also introduces bias, gut feelings, and legacy processes that aren’t always rational or efficient.

That’s why we need to be vigilant in distinguishing true systemic challenges from human bias—eliminating the latter wherever possible.

In an era of rapid technological evolution, it’s difficult—if not impossible—to make definitive statements about what will always require human intervention. The landscape is shifting too fast. That’s why we must constantly challenge ourselves, evolve our processes, and rethink our ways of working—to make the most of new technologies and overcome the limitations of the past.

Automation is powerful. But how we wield it—and where we let go of old habits—is what will define the future of media.

Author: Paolo

Economist by education, marketer by profession, coffee roaster by hobby.