People buy more soup when it's cold. This is not a sophisticated insight. It's barely an insight at all. And yet Campbell's Soup Company has been refining its ability to act on that simple truth for seventy years, across every evolution in media, and the story of how they did it tells us something important about what the agentic era actually disrupts.
In the 1950s, it was radio. Cold snap coming to Cleveland? Run the Campbell's spots. In the '60s and '70s, it was local television — same logic, bigger reach. By the 2000s, they'd built a programmatic weather-trigger system: when the National Weather Service forecast showed temperatures dropping below a threshold in a given DMA, digital ad budgets automatically surged. Warm front hitting Phoenix? Spend pulls back in real time.
They turned a grandmother's intuition into a science. And they were very, very good at it.
The sophistication compounded over decades. They built proprietary models correlating temperature gradients — not absolute temperature, but the rate of change — with demand lift by product line. Chicken noodle responded differently than tomato. Chunky responded differently than condensed. Regional preferences layered on top: New England clam chowder surges in the Northeast look different from cream of mushroom patterns in the Midwest. They knew things about weather-driven food behavior that no other company on earth could replicate.
Seventy years of institutional knowledge. And I'm about to argue that much of it may not matter.
The Signal Problem
Campbell's weather strategy worked because of a specific causal chain: weather changes → humans feel cold → cold triggers craving → craving drives purchase → marketing at the moment of craving accelerates the purchase. Every link in that chain involves a human experiencing a sensation and making a decision based on that sensation.
When an agent manages the household's meal planning, the chain breaks.
Not because weather doesn't affect food preferences — it does, and it will continue to. But because the agent doesn't experience cold. It doesn't crave soup. It processes its principal's stated preferences, dietary requirements, weekly meal plan, and budget constraints, then evaluates available options through structured product data. If the human said "I like soup when it's cold," the agent will factor that in. If the human didn't — if the preference was always implicit, felt rather than stated — the agent has no signal to act on.
Campbell's knows more about weather-driven demand than anyone on earth. And that might not matter at all — unless they can make agents care about what they know.
This is the signal problem, and it haunts every brand that built its competitive advantage on understanding implicit human behavior. Campbell's didn't need customers to say "I want soup because it's cold." They just needed to be in the right place at the right moment when the craving hit. The marketing was the signal — the ad reminded the human of something they already wanted but hadn't articulated.
An agent doesn't need reminding. It needs explicit instructions.
From Weather Triggers to Context APIs
Here's the interesting part: Campbell's institutional weather knowledge isn't worthless in the agentic era. It just needs to be repackaged for a completely different consumer.
Imagine Campbell's building a context API — a service that agents can query that combines weather data with Campbell's seventy years of demand modeling. The agent managing a household's meal plan in Chicago checks the forecast: temperatures dropping fifteen degrees over the next 48 hours. The agent queries Campbell's API: "Given this weather pattern in this region, what products historically see the highest demand lift? What's the optimal pairing? What's the nutritional profile?"
Campbell's responds with structured data: chicken noodle demand correlates at 0.87 with this temperature gradient in the Great Lakes region. Here's the nutritional breakdown. Here's the current price at three fulfillment channels. Here's a suggested meal plan that incorporates soup as a component.
The agent evaluates this against the household's preferences, cross-references with the weekly meal plan, checks the pantry inventory, and either adds it to the shopping list or doesn't.
In this model, Campbell's isn't buying media impressions to trigger human cravings. They're providing decision-relevant data to agents that are already planning meals. The weather knowledge compounds — it's genuinely valuable — but it's delivered through a data service, not an advertising campaign.
The Media Spend Question
This raises a question that should make every CPG CMO uncomfortable: what happens to the media budget?
Campbell's weather-triggered media strategy wasn't cheap. Programmatic buying, creative production, attribution modeling, the whole apparatus of modern digital marketing — it added up to significant spend designed to intercept humans at the moment of craving. If those humans increasingly delegate purchase decisions to agents that don't see ads, that entire channel becomes less effective by the year.
I'm not arguing that Campbell's should zero out their media budget tomorrow. The transition is gradual, and plenty of humans still browse the grocery store aisle and respond to advertising triggers. But the trend line is clear: as agent-mediated purchasing grows, the ROI on traditional media — including weather-triggered programmatic — declines.
The reallocation question is the hard one. Do you shift budget from media to data infrastructure? From creative production to API development? From advertising teams to engineering teams that build agent-facing services? The answer is almost certainly "some of each, gradually," but the organizational politics of that shift are brutal. Try telling a CMO's advertising team that their budget is being redirected to build a context API that agents can query. Try explaining that to the board.
But the math will force the conversation eventually. When the weather-triggered campaign that used to generate measurable demand lift starts showing diminishing returns because fewer humans are making impulse grocery decisions, someone will have to explain why — and propose an alternative.
The Implicit-to-Explicit Gap
The deepest lesson from the Campbell's story isn't about weather or media spend. It's about the gap between implicit preference and explicit instruction.
For seventy years, Campbell's succeeded by understanding what people wanted without being told. The weather-craving link was implicit — humans didn't articulate it, and they didn't need to. The marketing system anticipated the desire and met it at the right moment. That's actually a beautiful thing. It's marketing at its best: understanding people so well that you can serve their needs before they fully recognize them.
Agents operate on explicit instructions. "Buy soup when it's cold" is an explicit preference that an agent can act on. "I just feel like something warm and comforting" is an implicit desire that an agent struggles to interpret without context. The gap between those two — between the felt experience and the stated instruction — is where an enormous amount of consumer behavior lives. And it's where brands like Campbell's have built their entire competitive advantage.
The companies that figure out how to help customers translate implicit preferences into explicit agent instructions will have a massive advantage. Maybe that looks like an onboarding flow: "Do you like soup when it gets cold? Should I add it to your meal plan when temperatures drop?" Maybe it looks like the agent learning from behavior over time: the customer manually adds soup to the cart three cold weekends in a row, and the agent infers the pattern. Maybe it looks like Campbell's itself providing the translation layer — a service that helps agents understand the weather-mood-food connection that humans feel but don't articulate.
However it happens, the brands that bridge the implicit-to-explicit gap will be the ones that preserve their institutional knowledge advantages in the agentic era. The brands that wait for agents to figure it out on their own will lose to whoever makes the agent's job easier.
Campbell's spent seventy years learning that people want soup when it's cold. The next seventy years will be about teaching agents the same thing.