Asro
While the marketing industry remains fixated on the flashy promise of generative AI—the instant creation of ad copy, imagery, and video—a more substantive, albeit less "sexy," revolution is taking place beneath the surface. For global confectionery giant Hershey’s, the true competitive advantage of artificial intelligence isn’t found in a chatbot; it is found in the rigorous, data-heavy architecture of Media Mix Modeling (MMM) and the deployment of agentic AI systems that are fundamentally altering how the brand allocates its massive advertising budget.
The Paradigm Shift: From Reach to Relevance
For decades, the holy grail of advertising was reach—the pursuit of the lowest possible cost-per-thousand impressions (CPM). It was a game of tonnage. However, as digital ecosystems have fragmented and consumer attention has become increasingly scarce, Hershey’s has pivoted toward a more nuanced philosophy: "relevance over reach."
This shift is not merely semantic; it represents a fundamental change in the measurement ecosystem. Vinny Rinaldi, VP of consumer connections at Hershey, argues that the industry’s obsession with cheap reach often leads to a hollow outcome. "You can gamify any system to go after tonnage for a lower cost," Rinaldi notes. "It doesn’t mean you’re standing out."
Under this new framework, the guiding question is no longer "How many people saw this?" but rather, "Who was this truly right for at this exact moment, and did it change their behavior?" By prioritizing the quality of the connection, Hershey’s is leveraging AI agents to parse vast datasets, allowing the company to identify high-intent audiences and deliver messaging that resonates rather than just registers.
A Chronology of Transformation
The journey toward this AI-driven infrastructure did not happen overnight. It was the result of a deliberate, multi-year strategy to reclaim control of the brand’s data and technological destiny.
- 2018: Taking Ownership of the Stack. Hershey’s initiated a strategic move to own its direct contracts with major platforms, including a pivotal direct relationship with The Trade Desk. This ensured that while agencies would execute the buying, the brand would retain absolute transparency and access to the underlying data.
- 2021–2023: Customizing the Bidding Ecosystem. Hershey’s began a deep-dive collaboration with firms like Chalice to develop custom bidding algorithms. This period focused on integrating proprietary data sets into the buying platforms of YouTube, Meta, and others, laying the groundwork for more sophisticated automation.
- 2023–2024: The Agentic Infrastructure Year. Over the past 12 months, the brand focused on building a system of AI agents capable of providing media modeling in weeks rather than months. By partnering with tech-forward measurement firms like Mutinex and Tracer.tech, Hershey’s moved from static, retrospective reporting to dynamic, predictive modeling.
- 2024–Present: The "Upfront" Evolution. Currently, Hershey’s is utilizing its new agentic capabilities to inform its approach to the TV upfront season, using predictive revenue modeling to decide exactly where to pivot dollars as they diversify away from traditional cable.
Supporting Data and Strategic Outcomes
The efficacy of this approach is best exemplified by the brand’s recent, highly successful foray into Reddit. While many blue-chip advertisers view Reddit as a high-risk, "uncharted" environment, Hershey’s used its AI-driven measurement framework to validate the platform’s potential for its Reese’s brand.
The modeling revealed that Reddit—due to its hyper-focused community structure—was generating more units sold per dollar spent than almost any other channel in the company’s portfolio. The data-backed confidence allowed Hershey’s to double down on the platform, moving beyond generic display ads to more integrated, community-specific messaging.
This success is part of a broader shift in the brand’s media mix, which currently sits at approximately 72% digital and 28% traditional. The "machine"—the internal AI agentic infrastructure—now serves as the primary advisor for the brand’s treasury, allowing leadership to forecast the potential revenue impact of shifting dollars from legacy cable television into emerging digital channels.
Official Perspectives: The Role of Human Intuition
Despite the enthusiasm for automation, Rinaldi remains cautious about the limits of technology. He emphasizes that the "agentic layer" is meant to augment human intelligence, not replace it.
"I still think there’s human intuitiveness that’s needed," Rinaldi says. "Critical thinking is one of the most important levers that we have not focused on as an industry in general."
When asked about the possibility of AI agents eventually negotiating and buying media directly from publishers, Rinaldi expresses skepticism. He believes that the complexity of the media landscape requires a human touch to navigate nuances that a machine might miss. Instead, he views agents as tools to free up agency teams. By automating the grunt work of planning and buying, the brand’s agency partners can focus their time on the high-level strategy and creative thinking they are uniquely qualified to provide.
The Question of Ownership: Who Controls the Brain?
A critical theme in Hershey’s strategy is the insistence on brand ownership of the "agent." In the current agency-of-record model, where Publicis Groupe handles Hershey’s U.S. media duties, the temptation is often to let the agency build the tech. Rinaldi disagrees.
"To make it differentiated for a company like Hershey, inherently, we should own it," he asserts. Because the agents are fueled by proprietary sales data—information an agency does not own—the intellectual property of the model must remain with the brand.
This creates a triangular relationship: the brand owns the data and the contracts, the tech providers (like Mutinex or Tracer.tech) provide the infrastructure, and the agency (Publicis) sits in the "buying seat" to execute the strategy. This separation of duties ensures that if an agency relationship changes, the brand’s competitive advantage—its custom, data-fed AI engine—remains intact.
Implications for the Future of Marketing
The most important lesson from Hershey’s year of "agentic" development is the value of the unsexy, invisible work. Many companies have rushed to implement consumer-facing AI without building the necessary data pipelines, leading to what Rinaldi calls a "beautiful house built on a foundation that hasn’t had the concrete poured."
For the wider industry, the implications are profound:
- The Death of the "Black Box": Advertisers will increasingly demand that their media buying technology be transparent and proprietary. The days of relying solely on agency-provided, opaque black-box tools are numbered.
- Planning Over Buying: While the industry has been obsessed with buying efficiency (the "how"), the real value is shifting to planning (the "what" and "why"). AI agents will allow marketers to simulate thousands of scenarios before a single dollar is spent, turning planning into a rigorous science.
- The Necessity of Data Hygiene: The success of Hershey’s model is predicated on clean, integrated, and accessible data. Companies that have not modernized their data architecture will find themselves unable to effectively deploy agentic technology.
- Redefining Agency Value: Agencies that focus solely on buying will be marginalized by automation. Those that pivot to provide strategic, high-level guidance—utilizing the time saved by AI to perform deep critical analysis—will become more valuable than ever.
As the marketing landscape continues to evolve, Hershey’s has provided a roadmap for the modern advertiser. The future of media isn’t just about faster buying or cooler creative; it is about building a robust, proprietary engine that can translate data into actual business outcomes. For Hershey’s, the storm of change is coming, but because they have invested in their foundation, they are ready to weather it.
Nana
Jia Lissa
