May 18, 2026
Sweet deal or sugar meltdown?
Hershey Bets on Agentic AI to Rethink $2B in Marketing Spend
Hershey brings in AI to move ad money faster, but the internet says fix the candy first
TLDR: Hershey is using AI tools to make faster decisions about more than $2 billion in ad spending across its brands. Commenters were savage, saying no amount of smart software can fix candy people think tastes bad, turning a marketing story into a full-on product roast.
Hershey says it’s having a big main-character moment with artificial intelligence: the candy giant wants to use always-on AI tools to rethink more than $2 billion in marketing and trade spending, instead of waiting months for old reports to tell them what already happened. In plain English, the company is trying to make faster decisions about where to put ad money for brands like Reese’s and Skinny Pop. The pitch is simple: less spreadsheet lag, more real-time tweaking, and maybe 4% to 5% more revenue from media.
But the comment section? Absolutely not ready to clap. The loudest reaction was a brutal one: why optimize the ads when so many people think the product itself is the problem? One commenter went straight for the throat, saying Hershey should spend the money making chocolate that "doesn’t taste like vomit," while another argued the real return on investment would come from making products that don’t "taste like shit." Ouch. The vibe was less “wow, futuristic” and more “you can’t algorithm your way out of bad candy.”
Then came the AI cynicism. One of the sharpest jabs mocked the line that companies have a "data readiness problem," joking that firms don’t want real customer feedback and would rather "hallucinate the data they wanted." There was even a nostalgic drive-by from the Mad Men crowd: "They should have just listened to Don Draper’s pitch." So yes, Hershey wants a marketing revolution — but online, the real brand measurement seems to be how hard people can roast the chocolate.
Key Points
- •Hershey is using Mutinex and Tracer to automate marketing mix modeling and make marketing measurement faster and more frequent.
- •Mutinex’s system, supported by Claude and Gemini, is intended to provide an always-on MMM capability for monthly decisions across media and trade spend.
- •Tracer cleans and standardizes fragmented marketing and retail data so Mutinex’s models can run more quickly and reliably.
- •Hershey says the new setup can run models in as little as three weeks, compared with much slower historical analysis cycles.
- •The company is expanding MMM from about three analyses per year for roughly five brands to monthly measurement across its broader portfolio, and expects a 4% to 5% increase in media-attributable revenue.