Tech companies are churning out agentic AI tools, but agencies are still scratching their heads

Super Bowl viewers who were able to peel their eyes away from the spectacle of a brunette Guy Fieri in the Bosch ad may have noticed a trend in the commercials this year. Not that one— a lot of them were from AI companies.

With 125 million people tuning into the AI Bowl this year, the underlying push from marketers and ad agencies is relatively simple: follow the eyeballs. AI agents are top of mind for brands and agencies because consumers are using the technology more and more. What makes this technological advancement different from previous technological industry moves, though, is that both the industry and the consumer are awash in confusion.

As platforms flood brands and consumers with agentic AI tools, media buyers are navigating not only a nascent but rapidly evolving technological landscape, but also varied client sentiment toward agents and perhaps, AI in general.

“On one hand, we see a major rise in conversational commerce where people are describing what they are looking for in the same way as if they’re talking to a friend or talking to a sales associate. The queries are getting much longer, much more descriptive,” Ashish Gupta, VP and GM of merchant shopping at Google, said. “At the same time, what you’re also seeing is that agentic technology is finally catching up, which means that shoppers can delegate some of the most serious part of shopping to the agents…But for this new agentic era to be really and truly successful, it has to work for everyone, especially all the brands and retailers of all sizes, which are the backbone of our industry.”

Talk to me

Recently announced agentic tools from major tech platforms Google and Meta are focused on commerce, with the expectation that users will become more exposed to shopping via agents. Google made a number of related announcements at the National Retail Federation annual show in January, including the debut of the Universal Commerce Protocol (UCP).

UCP, developed with companies including Walmart and Shopify, provides a way to standardize and streamline agentic commerce agents, and is being integrated across Google features like AI Mode in Search and in the Gemini app. According to the company, its checkout feature enables “consumers to go from discovery to purchase seamlessly,” using existing payment information stored within Google Wallet.

Salesforce is already integrating the standard into its agentic platform, Agentforce Commerce, and it’s being used to support clients in a few different ways, such as making sure that UCP is given up-to-date product data so it won’t make an error like suggesting an item that isn’t in stock.

Retailers can also use Google’s newly debuted Business Agent to create branded AI agents which act as “virtual sales associates” that can answer consumers’ questions, according to a Google press release.

“We are not a retailer,” Gupta told Marketing Brew. “We are not a marketplace. How we think about ourselves is [as] a matchmaker.”

Gupta claims brands are seeking to be “more discovered in conversational AI,” which he said can be achieved if they provide Google with detailed product data.“These attributes allow the brand to give us even more rich descriptive details about the product, like common answers and questions that people are asking about those products or what the compatible accessories are which go along with the product,” he said.

To that end, the company is rolling out additional data attributes in its Merchant Center that let brands input “richer descriptive details about the product,” Gupta said.

“Let’s say a shopper comes to AI Mode on Google Search, and they are looking for a suit for a summer wedding, and they want something which is lightweight, something breathable,” Gupta said. “When they are searching for this on AI Mode, we are allowing merchants to give us typical questions and answers which customers may be asking for.”

Google’s data attributes rollout is part of a larger shift happening, in which LLMs need substantial information on a brand to help it surface in search results, Chuck Gahun, Forrester principal analyst, said. Reddit and user product reviews are just some of the forms of information LLMs are consuming to determine how to populate answers or search results for user queries; brands are responding by publishing more quality product content, he said.

“They’re going after the content everywhere to make sure it’s holistic, so that wherever Google’s LLM or OpenAI’s LLM is learning from, it’s accurate and comprehensive,” he told Marketing Brew.

While Google’s new data attributes will initially only roll out to select retailers, some brands are already thinking more widely about how to connect agents with detailed product data to capture consumers’ various niche queries. Travel company Booking Holdings, which owns sites including Booking.com and OpenTable, is connecting its in-platform agents to data on properties represented on its platforms to give agents information that can help meet extremely specific queries, Leslie Cafferty, chief communications officer at Booking Holdings, said.

“‘Is the restaurant in Milan dog-friendly?’ is not going to be a mass query,” Cafferty said. “But we use agentic capabilities to interact with customers…to go match that [query].”

Shopify is similarly aiming to bring brands and consumers closer through agentic commerce, Amanda Engelman, Shopify’s director of product for merchant marketing, said. In December, Shopify debuted Agentic Storefronts, which lets merchants surface their products on AI platforms like Perplexity and ChatGPT.

“At a very high level, our approach to agentic commerce is really through the path of, ‘How do we enable people to move quicker?’” she said. “Whether that’s through distributing their products in places where people are spending time, like on ChatGPT, as an example, or if it means using AI…to help them source better or just be smarter about their decision-making,” she told Marketing Brew.

Slow and steady

While tech companies churn out agentic tools, agencies and media buyers are at varying degrees of buy-in. Earlier this year, WPP rolled out several agents, including a brand analytics agent. When it comes to media buying, others, like Butler/Till, are exploring using AdCP, the open standard rolled out in October by Scope3 and other ad-tech leaders to bring agents to media buying and other advertising tasks. Butler/Till ran a campaign for beverage company Clubtails involving AdCP partner PubMatic’s Activate Platform to execute agentic buying in December. The agentic campaign “delivered more than 5x cost efficiencies,” PubMatic CEO Rajeev Goel said on the company’s Q4 2025 earnings call in February.

According to a recent Digiday article, Butler/Till’s agentic media buying campaign for Geloso Beverage Group “successfully cut intermediary fees by over 80% and reduced CPMs while hitting industry benchmarks on fraud and inventory standards, according to the agency and its test partner PubMatic.”

Media buyers still have a ways to go to establish an agentic buying framework, however. A January IAB report noted that 66% of agencies surveyed are focusing on agentic AI for ad buying and campaign execution, but 40% of respondents listed the need to understand its use in those areas as a challenge.

Valerie Renda, AVP of business transformation at Kepler, echoed similar sentiments. The agency is monitoring the evolving agentic media buying space and may conduct more tests down the road. Kepler is also seeing varied client sentiment around agentic AI use, with some more apprehensive than others. Those in highly regulated industries, like finance and healthcare, are among those she expects to not be especially keen. (A faulty financial transaction or diagnosis can hurt brand credibility; data privacy is another concern.)

However, that hasn’t stopped Kepler from establishing a new business transformation unit to figure out, in part, how to incorporate agentic workflows across the agency.

“When it comes to agentic AI…we need to have ownership, we need to have accuracy, and we need to have accountability for anything that AI or agents are producing,” she said. “In order to do that, we need a governance model, and that’s what my department is in charge of.”