Artificial intelligence (AI) has reached marketing faster than almost any previous technology shift. Within months, generative tools moved from experimentation to expectation. Yet despite unprecedented adoption, AI has not made most marketing organizations meaningfully smarter.
A consistent pattern becomes difficult to ignore: AI has entered marketing at scale, but it has not entered with a strategy. Instead of producing a durable competitive advantage, it is flattening it. The industry now faces a risk of competitive mediocrity at machine speed.
It was against this backdrop that madSense, the AI-driven AdTech company, began positioning itself not as another AI tool provider, but as a response to a deeper structural problem inside modern agencies. Rather than accelerating existing workflows, madSense set out to rebuild how agencies learn, decide and retain intelligence.
“From the very beginning, we’ve been set on a deliberate and disciplined path to help brands and agencies walk strongly into AI transformation within their own organizations,” Gordan Topalovic, CEO of madSense, told me. “Everything we’re doing is informed by more than 15 years of experience building martech platforms, publisher-facing products, large-scale audio deployments, and CTV systems, and asking how AI can finally be used in a way that produces consistent, explainable, and meaningful outcomes.”
When Speed Stops Being an Advantage
Thanks to AI, most agencies can now generate content faster than ever. But as similar tools spread across the market, outputs begin to converge, campaigns feel increasingly interchangeable, and insights often read as lightly remixed rather than truly original. Executives increasingly describe spending more time managing AI systems than eliminating work, as tool sprawl quietly becomes a new operational tax.
The problem today is not access to AI; it is coherence.
“Marketing’s AI problem has never been about access to tools. The industry is saturated with tools. The real issue is structural thinking—how to extract a signal that a campaign planner or researcher can actually rely on. Since the dot-com era, we’ve seen a compounded explosion of data from disconnected systems, and AI is accelerating that growth even further,” Topalovic explained. “Information exists within that data, but it’s fragmented, siloed, and overwhelming, and there is only a finite amount of time and human capacity available to interpret it responsibly.”
Agency leaders consistently describe three paths forming across the industry. Some organizations have become AI-enhanced, layering generative tools onto existing workflows. These agencies move quickly and show early productivity gains, but they retrofit AI into systems never designed to learn or retain context.
Others push further and become AI-enabled. They integrate AI across research, media, and creative functions, unlocking real efficiencies. But decision-making remains fragmented. Intelligence does not compound. These agencies often find themselves stuck between progress and reinvention—productive, but structurally constrained.
Likewise, a far smaller group is rebuilding as AI-first agencies. These organizations treat AI not as a productivity layer but as an operating infrastructure.
“If we follow the same path marketing has followed for the last 15 years, we will almost certainly fail. That path has been defined by layering new tools on top of broken workflows,” Topalovic told me. “Our approach starts from a different assumption: AI should be used to extract reliable signals from disparate data, while still relying on human experts (campaign specialists, brand managers, domain leaders) who understand their field better than anyone else and know exactly where to look, what matters and what should be ignored.”
Why AI Pilots Keep Failing Quietly
Industry research consistently shows that 95% AI initiatives stall at the pilot stage, failing to deliver any measurable business returns.
“More than three quarters of marketing leaders openly admit that significant portions of their budgets are being wasted on AI initiatives that don’t deliver meaningful results,” Suresh Kumar, board advisor at madSense, told me. “That gap between expectation and reality is creating frustration, skepticism, and a growing lack of trust in AI as a whole.”
Off-the-shelf AI agents often deepen the problem. Many function as polished chatbot wrappers, producing passable outputs while pulling proprietary data into opaque systems.
“Most providers are taking generic models like ChatGPT or Gemini, layering them over old processes, and calling it innovation. That’s great for writing emails or summarizing meeting notes, but it doesn’t help us decide how to deploy multimillion-dollar media budgets with confidence,” said Brandon Heagle, chief digital officer at Stella Rising. “If leadership teams can’t understand how a recommendation was generated, then they won’t act on it with confidence. We needed a partner like madSense because we needed logic we could actually see, audit, and trust. We needed an engine that understands bespoke decision-making, not just text prediction.”
For Topalovic, trust in AI begins with traceability. Lack of visibility, he argues, sits at the core of why many leaders remain uneasy about AI today.
“Most commercial AI platforms struggle to explain how they arrive at an answer. That’s not a theoretical critique—it’s direct feedback from clients who tested these tools themselves. When you ask them what data was used or how it was combined, the answers become vague or incomplete,” he explained. “The real differentiator is fidelity: the ability to replicate processes, trace every data input, and provide complete transparency into the reasoning behind each output. Without that, AI remains something you experiment with, not something you trust at scale.”
Solving for trust, in other words, requires more than better models—it requires an architecture that can connect intelligence end to end. madSense’s recent launch of its Intelligence OS is designed as a direct response to that problem. Rather than locking users into proprietary logic, the system is designed to expose how signals are interpreted, how decisions are formed, and how outcomes evolve over time—restoring a sense of agency that many marketers feel AI has taken away.
Heagle added that traditional adtech is great at chasing the cheapest click or the lowest CPM. But in beauty and wellness, the cheapest click is rarely your best customer.
“madSense allowed us to feed in messy, fragmented signals—everything from merchandising data to behavioral trends—and it reasoned across them to find the pockets of profitable growth. It is also fundamentally changing our team’s workflow. Rather than spend time wrangling data and manually reconciling spreadsheets, the platform helps automate work and speeds analysis, which meant our strategists could actually be strategists.”
The launch marks the first AI stack capable of addressing the $26.8 billion in wasted ad spend that continues to plague the programmatic ecosystem.
“We don’t enter organizations selling products; we enter selling outcomes. That means starting with where the biggest inefficiencies exist today, proving value incrementally and mapping transformation across 30-day, 60-day, 90-day, and longer-term milestones. We’re very explicit with clients that this is a journey, not a transaction,” Suresh said.
From Execution to Orchestration
As AI lowers the cost of execution, clients are reassessing what they value from agencies. Speed and volume matter less when machines can generate both instantly. What remains scarce is judgment, foresight and systems that learn over time.
“AI adoption isn’t optional, agencies will have to adapt to stay relevant. But the balance matters. AI is here to stay, yet how deeply it’s embedded is ultimately a human decision, not an AI one,” Topalovic said. “Until we can rely on AI in the same way we rely on each other: as a trusted teammate that can be questioned, corrected, and understood—we will always be better positioned when human expertise remains central to the system.”
Marketing’s AI transformation will not hinge on the next model release or the loudest launch. It will hinge on whether organizations can build systems that compound knowledge, explain decisions and deliver measurable business impact.
For agencies willing to rebuild from the ground up, AI offers an asymmetric advantage. For those content to layer new tools onto old logic, the danger is quieter—but far more enduring.
