A CEO’s Perspective on the Future of AI in Programmatic
AI is now the dominant narrative across advertising. But long before generative AI entered every conversation, machine learning had already become foundational to programmatic infrastructure.
Having built and scaled adtech companies for over a decade, and now leading Kayzen, Tim Koschella has witnessed programmatic evolve from early automation systems to today’s AI-infused ecosystems.
In this conversation, he shares his perspective on where AI truly adds value, where it is overstated, and why human accountability must remain central as systems become more intelligent.
What’s Actually New and What Isn’t
While AI dominates headlines today, Tim grounds the conversation in history.
“Let’s establish a fact first: AI has been at the center of adtech for more than ten years already. Advertising is one of the earliest markets where AI was applied at scale.”
What most people now call AI has long been part of programmatic advertising, where machine learning systems operate at a massive scale. These systems rely on structured data to predict outcomes, whether a user is likely to click, convert, or purchase, and then return those predictions within milliseconds.
Generative AI changes the interface layer. “It’s more flexible. It handles unstructured data. It tolerates imperfect inputs. It’s a much better companion for humans.” That shift, however, does not make it universally superior. “In RTB, speed matters. Inference calls need to run within milliseconds. Generative AI often needs seconds. That time lag makes it less useful for machine-to-machine communication.”
Usefulness, in his view, is practical rather than philosophical. “Anything that improves measurable outcomes for advertisers or helps humans do their job faster and better is useful. As long as you put the customer problem first, and find ways AI can solve it, it’s a path worth pursuing.
Why Performance Is Still Hard to Explain
Despite AI being embedded in advertising infrastructure for years, many teams still struggle to clearly explain performance outcomes. The issue, he argues, is less about model sophistication and more about definition. Performance, he emphasizes, has to start with a deep understanding of the advertiser’s business and the goals the company aims to achieve with its marketing investment.
“At its core, performance is used in so many different ways that it remains vague until it’s clearly defined in context. A brand agency managing Coca-Cola’s mobile ad budget in Europe and a mobile RPG game looking for growth in Japan will likely have a very different perspective on how to define and measure performance.”
Optimization conversations often stop at the metric label rather than the business reality behind it. “Take ROAS as an example. When someone says they optimize for ROAS, I immediately ask: which ROAS? D7? D30? Lifetime? How does your product generate revenue? Are you driven by outliers? How do you measure incrementality versus attribution, and what do you do when they point in different directions?”
The complexity is not purely technical; it is strategic. “We need smart professionals to make sense of this complexity. AI can help, but humans have to stay at the center of decision-making.”
The Pressure to Have an “AI Story”
If AI has been embedded in advertising systems for years, why does the current moment feel so amplified? Much of it, Tim suggests, comes down to narrative pressure, particularly from investors and capital markets.
“The pressure to include AI everywhere is tremendous, especially for companies backed by large investors or listed publicly. Every CEO is expected to have an AI equity story.”
He does not dispute that AI can improve most businesses. What concerns him is how it is framed. “AI doesn’t have to be the fabric that connects everything in your business, like the tissue in your body. In many cases, it’s simply a useful tool to improve efficiency and quality.”
When AI becomes an identity rather than a capability, expectations can quickly outpace reality. “If those intentionally created expectations are not being met down the road, it may create a sentiment of delusion. And those companies which actually have value-generating applications of AI may need to make an extra effort to explain themselves.”
He does not argue against AI adoption. What concerns him is the inflation of narrative beyond practical value. In many cases, he notes, AI is a tool to improve efficiency and quality, not a reinvention of how businesses operate. The differentiator, then, is not how aggressively AI is introduced, but whether it solves real customer problems and delivers measurable value.
Principles Before Technology
That concern leads to a deeper question: if AI systems are becoming more capable, what should guide their use?
“In programmatic advertising, there are many data signals that AI can process, interpret, and act uponI. But there is also noise in the signal that can lead to false or misleading interpretations. AI can easily be misled.”
At the scale of the global programmatic ad market, the consequences of misinterpretation can be severe. “A wrong action can lead to losses that can render entire companies bankrupt overnight. It’s similar to financial markets. If you were to automate financial transactions without imposing a system of limits and checks and balances, you could just blow up the whole thing in one day.”
Human oversight, therefore, is structural rather than optional. “AI will never replace the critical thinking and decision-making humans bring to the table. Ultimately, the responsibility and accountability for the actions of the AI we use lie with us humans.”
There is, however, a distinction between interpretation and execution. “Few people are actually good at interpreting large, complex datasets. So the chances that a human will ‘get it wrong’ are likely much higher than the AI getting it wrong.” But when automation directly controls capital without guardrails, the risk profile changes. “If AI gets it wrong, it may spend an unapproved $1M budget in the wrong place faster than you can blink. These kinds of major high-impact mistakes are less likely to be made by humans.”
Where AI Is Underused
Where should AI create the most value before money is spent?
“In programmatic, that’s clearly the planning phase of a campaign. Today, AI is very active during the optimization phase. But AI is rarely used in pre-campaign planning, where the strategy is decided, and the expectations are set.”
Intelligence has been disproportionately focused on live campaign execution, while the strategic phase, where assumptions are formed and expectations are set, remains comparatively underdeveloped. Before execution begins, AI can help teams model scenarios and make systematic use of structured first-party signals. “The most valuable asset before the first money is spent is advertiser first-party data, data about their existing users, funnel metrics, seed audiences for lookalikes, and related segmentation inputs. This is where systematic use of AI can add a lot of value without requiring campaigns to spend money.”
Once campaigns are live, intelligence should extend beyond acceleration and efficiency to support understanding. “AI should help teams learn rather than just react. It should surface anomalies and highlight trade-offs.” As systems become more capable, clearly defined guardrails become even more important. “For any major decision that operates outside of defined boundaries, humans should serve as the final decision layer.”
The Outcome: Empowerment
The conversation ultimately shifts from infrastructure to mindset. Tim challenges one of the industry’s core assumptions: “One common myth is that you need a huge team of data scientists to compete with big incumbents. What matters more is the quality of your team members and the way they collaborate to establish a strong understanding of the problems that need to be solved.” Alignment across disciplines — data science, engineering, and business stakeholders — matters more than scale alone.
He is equally cautious of exaggerated technical mystique. “Beware of small companies that claim their smart team of scientists and ex-NASA engineers have built solutions no one else ever could come up with. It’s likely a nice sales pitch.”
If AI is deployed thoughtfully, marketers should not feel replaced. They should feel capable of doing more with clarity and control. “Nowadays, a team of five can easily manage a $500M annual marketing budget while still being in control of making real-time campaign flight adjustments and optimization decisions every single day.”
For Tim, intelligence should scale capability without eroding responsibility. “AI should empower decisions. It should never replace judgment.”
