For the past year, Matt Stoller and I have been co-hosting Organized Money, a podcast about all the money and power in the world. And I think the whole project has been leading up to the episode we released on Friday, about Google’s fairly advanced effort to organize the world’s pricing through their AI models.
At one level, it sounds relatively benign: Google is partnering with a bunch of payment systems and big-box stores to offer shopping assistants to shoppers. We are now familiar with these chatbots (or in the new annoying buzzword of the day, “agentic” commerce) in our daily lives; you must consult one if you go to customer service to complain about a product, since you pretty much can’t find a human being anymore. What is the harm of creating a similar chatbot to help you with shopping?
But that’s before you actually find out the plans that Google has in store, and the ambitions to become the pricing agent for as many merchants as possible. The possibilities here are endless: Google determining prices for multiple big players in the same product category; Google using its vast wealth of personal information (along with what its partners know) to pinpoint a customer’s willingness to pay at the highest possible price; Google monopolizing AI through these shopping tools just as it monopolized the platform economy in Web 2.0, and much more.
You can listen to the whole episode, which also features Lindsay Owens of Groundwork Collaborative, who was among the first to notice Google’s plans. But here’s a brief description of what’s going on.
Earlier this month at the National Retail Federation’s annual conference, AI was all the rage. Pricing consultants and advisers have sold the industry on the idea that their tools can maximize profits in all phases of the business. And a keynote speaker was Google CEO Sundar Pichai.
Pichai announced that Google had created something called the Universal Commerce Protocol, or UCP. At root, it’s a chatbot: a conversational AI agent that will interface with customers to help them find what to buy. Of course, Google has all the existing data on those potential shoppers: email history from Gmail, web search history from Google Search, viewing habits from YouTube, payments from Google Pay, and much more.
It’s like walking into a store where the sales clerk has a perfect graph of your wants, needs, and desires, and perfect information about everything you’ve browsed before, everything you’ve talked about with your friends, and maybe a lot of things you’ve bought. And throughout its UCP announcement, Google stresses how it will obtain broad consent to access this information from customers and share it with others.
But Google’s vast storehouse of information gets supplemented here. Google is partnering on UCP with Shopify, Etsy, Home Depot, Gap, Wayfair, Target, and Walmart, so presumably all of your shopping history from those sites will be ported into this agentic AI commerce bot. Often these big-box stores have “social graphs” of their customers enhanced with third-party data to build out the profile. Google will “power” these chatbots, but they will be branded with the particular company, so customers will be unaware that their email or YouTube history will be part of the mix determining how they are steered.
In addition, Google is partnering with American Express, Mastercard, PayPal, Visa, and Stripe, the companies where people make the vast majority of all of their noncash purchases, especially online. That sales clerk, then, has access to an instant record of practically all of your retail history. Plus, Google just made a deal with Apple to have its Gemini AI model power Siri on all iPhones and Apple products. That adds in the information from every phone in the world, practically speaking, since Android phones are already powered by Google.
That information can then be used to present you with prices specifically targeted to you. Google is very careful to call these “discounts,” because discounts are typically exempted from any bans on deceptive practices in pricing. But if Google is setting the reference price, any discount off that isn’t really a discount at all; it’s just a differential price based on circumstances. And indeed, “dynamic pricing” is listed as one of the attributes Google touts in UCP.
So this is really the mother of all surveillance pricing schemes. There are “cross-sell” and “upsell” modules as part of UCP that are intended to maximize the amount spent over time. “Direct offers” is another tool that determines which specific options to buy a particular customer gets to see. One of the goals, stated explicitly in the announcement, is to “help shoppers prioritize value over price alone,” using its information and behavioral tools to advance that. And even though the bot will feel like it’s working for the customer, in reality it will be working for Google and its retail partner, in their interests.
Who knows exactly how Google will seek to profit from this? It might get a percentage of the upsell or of every transaction. Or it might charge for the partnership AI bot, and encourage every retailer to sign up by claiming that this is inevitable and that if you don’t get in bed with Google you’ll be left out. That’s especially true if Google wins the battle for AI and these agentic bots, which OpenAI at least is concerned about right now.
As this expands, you can see how this can trend prices upward for entire industries. If Walmart and Target’s agentic AI bots are both powered by Google, they will converge on price for particular individuals. That used to be called price-fixing, and it remains illegal even if an AI bot does it.
The real killer app of this, in my view, is when Google encourages shoppers to allow the AI bot to make regular semiautomatic purchases for them of everyday items they need. If the bot knows that you get batteries approximately every three months and orange juice every other week, they can make it so that just shows up on your doorstep, disconnecting the customer even further from price. The bot will feel like it’s working for the customer, but it’s partnering with the retailer, and setting the price.
This is a dangerous combination, in my view. All previous attempts at AI-fueled pricing have been met with absolute scorn from the public: Wendy’s digital menu boards, Delta Air Lines saying 20 percent of its fares would be generated by AI. But there’s a potential marketing pitch for this Google thing as a labor-saving device, as relieving you of the hassles of errands. Everything shows up right when you need it and you don’t have to think about it anymore. That could be a powerful selling point, masking the personalized pricing, the consolidation and price-fixing underneath.
Now, what’s interesting here is that Google felt the need to respond to Owens when she pointed all this out on Twitter. That’s very rare; Google usually makes considered announcements and does not act as a reply guy. Moreover, what they claimed—that direct offers will only allow discounts and “cannot be used to raise prices,” and that upselling merely refers to “premium product options that people might be interested in”—falls into the same trap of claiming “discounts” on reference prices that they themselves set. (It also is at odds with what it has said in current litigation, that it does nothing to restrain price.)
This is not a done deal, as Owens points out in the episode. Google Shopping through the AI tool right now is crude and kind of bad; she related it to what it would be like to do “online shopping in Microsoft DOS” in the mid-1990s. And the companies partnering with Google might get wise to Google trying to maximize its cut or make them dependent on AI shopping, and break off. Plus, because it’s in its embryonic stages, there’s still time to get policymakers and regulators involved in setting up guardrails.
It’s also a massive comedown from AI’s ambitions. We went from creating God in the machine and dominating life on Earth to trying to chisel a 10 percent upsell on a rug. This is something policymakers at least have the capacity to deal with.
I encourage you to listen to the whole episode. And we’re going to post all episodes of Organized Money at the Prospect, so you can have a bigger window into how the business world really works and what can be done to make sure it works for all of us.