Milena Traikovich is a seasoned expert in demand generation and performance optimization, known for her ability to cut through the noise of the rapidly evolving marketing technology landscape. With a background deeply rooted in analytics and lead generation initiatives, she has witnessed firsthand how the current wave of artificial intelligence is both a revolutionary opportunity and a potential minefield for advertisers. In this conversation, we explore her tactical framework for vetting AI vendors, the necessity of industry-specific expertise in software development, and the critical importance of maintaining data integrity in an automated world.
The following discussion explores the strategic evaluation of AI tools, focusing on the distinction between feature-heavy marketing and actual business utility. We delve into why a vendor’s personal experience in media buying can make or break a partnership, the calculated risks of being an early adopter versus choosing established platforms, and the non-negotiable standards for data ownership. Finally, we address the true cost of implementation, which often extends far beyond the initial price tag.
Many AI vendors lead with complex, feature-heavy language that can often mask a lack of practical utility. How do you look past the technical jargon to determine if a tool will actually solve a core business problem?
The marketing world is currently saturated with noise, where for every one smart initiative you want to launch, it feels like 10 AI vendors have suddenly appeared with a tool claiming to be the perfect solution. To see past the dazzle, I force the conversation back to specific business outcomes and the challenges my team faces every single day. If a vendor cannot clearly articulate how their features map to a real-world problem, such as increasing output or identifying gaps in tracking to speed up troubleshooting, it is a massive red flag. I want to hear a clear narrative about how the tool improves outcomes, backed by a case study from an organization of a similar size or vertical. It is easy to get caught up in promises of “saving time,” but I always tell my team that saving time is only valuable if we have a concrete plan for how to reinvest that reclaimed time into higher-value strategy.
You have mentioned that there is a significant difference between tools built “for” advertisers and those built “at” advertisers. What specific signs should a buyer look for to know if the developers truly understand the daily life of a media buyer?
Technical expertise is a given in this space, but it is the operational empathy that truly separates a great tool from a mediocre one. I look for a vendor who can describe the granular frustrations of a media buyer’s day, from the repetitive nature of manual bid adjustments to the anxiety of reporting on fluctuating lead quality. If the founding team doesn’t have personal experience in the trenches of media buying, they must be able to prove they have conducted extensive research to incorporate those professional insights into the UI and functionality. When a rep has a shallow understanding of the market, it suggests the tool might be clunky or irrelevant to my actual workflow. On the flip side, if the founder’s mission stems from a problem they identified while running campaigns themselves, that’s a compelling story that usually leads to a much more intuitive and effective product.
When considering a new AI solution, how do you weigh the potential “first-mover advantage” of being an early adopter against the risks of bugs and unproven results?
Falling into the early adopter camp is a high-stakes gamble that requires a very specific appetite for risk. You might find a growth accelerator that puts you miles ahead of your competitors, but you are just as likely to spend months spinning your wheels while working through technical bugs or realizing the tool simply doesn’t deliver on its hype. Transparency is my green flag here; if a vendor is honest about us being one of their first clients in a vertical, I can respect that and potentially move forward. However, if I’m going to be an early adopter and act as a de facto QA tester, the vendor needs to be incredibly flexible on contract terms to mitigate my financial and operational risks. For more established vendors, I expect to see specific, relevant case studies with real numbers that prove they can handle an organization of our scale without the growing pains of a startup.
Data privacy is a major concern in the age of large language models. What are the absolute non-negotiables you look for in a contract regarding how a vendor handles your company’s data?
It is honestly alarming how quickly some people are willing to hand over their proprietary data in the rush to gain a competitive edge. My stance is firm: you own your data, full stop, and that must be reflected in the written contract rather than just a verbal assurance from a salesperson. I look for explicit clarity on where the data is stored, how long it is retained, and exactly how it is being used to train models—specifically ensuring it isn’t being used to benefit third-party models or competitors. If the terms of service are vague or seem to contradict what the sales rep is telling me, I stop the conversation immediately. We need to know exactly what happens to our data if we decide to stop using the tool, and any refusal to put these protections in writing is an automatic nonstarter for me.
Implementation is often where the hidden costs of a new tool emerge. What should marketing teams consider regarding internal resources before they sign on the dotted line?
The real cost of a tool is almost always higher than the price listed on the invoice because it involves the “internal lift” of your own team. Before committing, you have to realistically assess the time required for integration, employee training, and ongoing quality assurance, as well as any potential disruptions to your existing martech stack. I have seen far too many marketers waste a significant portion of their budget on tools that end up sitting on the shelf because the team didn’t have the bandwidth to actually implement them properly. No tool is truly one-size-fits-all, and if the design isn’t intuitive enough for the team to adopt it quickly, the investment will never reach its full potential. If a tool seems too hard to onboard or requires resources we simply don’t have, I would rather wait for a more streamlined solution to emerge than force a square peg into a round hole.
What is your forecast for the future of AI in demand generation over the next year?
I expect we will see a massive shift from the current “hype cycle” of general AI tools toward much more specialized, vertical-specific solutions that prioritize deep integration over broad features. Right now, there are 40,000 marketing professionals and even more vendors trying to figure this out, which creates a lot of overlapping noise, but the next 12 months will likely see a consolidation where only the tools with proven utility survive. We will move away from being dazzled by the fact that a tool “uses AI” and start treating it like any other part of our stack, where the only thing that matters is the measurable impact on lead quality and customer acquisition costs. My advice for readers is to remain curious but cautious; if a tool feels too expensive or rigid right now, don’t rush—a more effective and user-friendly version is likely just a few months away.
