AI agents were everywhere around the Super Bowl this year—so brands are doing the obvious thing: chasing attention where the attention is. The twist is that both consumers and the industry are still figuring out what “agentic” actually means in practice, even as platforms keep shipping new tools. Google’s Ashish Gupta pointed to longer, more conversational shopping queries (people basically talking to search like it’s a sales associate), and argued that agents are finally good enough to take on serious shopping tasks—assuming it works for brands of every size, not just the usual retail giants.
On the platform side, the agent push is heavily commerce-shaped. Google introduced the Universal Commerce Protocol (UCP), built with players like Walmart and Shopify, to standardize how commerce agents work and connect across experiences like AI Mode in Search and the Gemini app—plus checkout that can use stored Google Wallet payment info. Google also debuted a “Business Agent” that lets retailers create branded “virtual sales associates.” The big operational takeaway: getting found in conversational AI is becoming a data discipline. Google is rolling out richer Merchant Center attributes (initially to select retailers) so brands can feed more structured product detail—FAQs, compatibility, descriptive nuances—because LLMs are pulling from many sources (including reviews and Reddit) to decide what to recommend. Booking Holdings is already wiring agents to its property and restaurant data to answer hyper-specific questions (think: “Is the restaurant in Milan dog-friendly?”), while Shopify’s Agentic Storefronts aim to place merchants’ products in AI destinations like Perplexity and ChatGPT.
Agencies, meanwhile, are cautiously testing instead of cannonballing in. WPP has rolled out multiple agents (including one for brand analytics), and Butler/Till tested agentic media buying using the AdCP standard with PubMatic’s Activate Platform for Clubtails—PubMatic says the effort drove “more than 5x cost efficiencies.” Digiday also reported that Butler/Till’s agentic buying test for Geloso Beverage Group cut intermediary fees by 80%+ and reduced CPMs while still meeting industry benchmarks for fraud and inventory standards. Still, the “how do we operationalize this safely?” questions are loud: an IAB report found 66% of agencies are focusing on agentic AI in buying/execution, yet 40% say understanding how to use it there is a key challenge. Kepler’s Valerie Renda flagged uneven client appetite (regulated categories especially), and emphasized the unsexy but essential next step: governance—ownership, accuracy, and accountability for what agents do.
Read more at Marketing Brew.