Good morning. Andrew here. In an era where digital agents will soon make purchases on our behalf, how do retailers persuade algorithms instead of humans? We asked Brent Crane, a DealBook contributor, to take us inside the emerging world of marketing to artificial intelligence. Also: Sarah Kessler has an excellent Q&A with our friend and former DealBook colleague, Kevin Roose, about his new book, “The AGI Chronicles.” It’s fascinating — and I’m not just saying that. Finally, don’t miss our news quiz — it should be a breeze for the Hollywood crowd, but a bit trickier for everyone else. (Was this newsletter forwarded to you? Sign up here.)
Marketing to bots
When an A.I. agent goes shopping, how does it decide what to buy? That’s the big question on every marketing executive’s mind as customers increasingly turn to chatbots for recommendations and as new A.I. agents from Meta and OpenAI offer to help anyone send a bot shopping on their behalf. A new category of start-up is vying to help brands figure it out. Straddling the line between tech and marketing, its strategies depart sharply from traditional advertising. “When you advertise to humans, you target eyeballs, emotions and dopamine, but when you advertise to agents, you advertise to logic, a very mathematical logic,” said Aviv Shamny, the co-founder and chief executive of Limy, a “AI search platform” backed by the venture capital firm Andreessen Horowitz. Not long ago, most websites blocked bots. Some stores, like Amazon, are also blocking some A.I. agent shoppers. But many retailers and brands want to cater to A.I. Traffic to American retail sites from A.I. sources increased 393 percent in 2026, according to a report from Adobe. Experts envision a more agent-centric future. By 2030, “agent-influenced spend” might constitute up to 20 percent of total U.S. e-commerce, or $385 billion, according to Morgan Stanley research. “More and more, you’ll just need to click ‘buy’ in the AI. chat interface — you’ll never go to a brand’s site,” said Jeff Blackman, the managing director at Barbarian Group, a marketing agency that is leaning into A.I. “The risk of not being prepared could mean you won’t show up.” Agentic commerceTyler Ackerman, the chief executive of the RTA Store, a family-owned online cabinetry retailer, decided to invest in his company’s A.I. visibility last year, after team members noticed that the company was not showing up in LLM recommendations anymore. “If we’re not focusing on, ‘How do we have our products read easily by these agents fast and securely?’ then we’re left behind,” Ackerman said. The RTA Store paid $15,000 to New Generation, a San Francisco start-up that promises to transform companies’ “static product catalogs into structured data and generative interfaces that A.I. can understand.” New Generation made a number of changes to Ackerman’s website, which he says have worked, from its basic code to adding an A.I. widget that fields product inquiries (from both human and non.) But it also added reams of content describing the intricacies of each product offering — something that A.I. marketing experts often suggest. Agents “read” more than human shoppers, preferring “content that’s very factual, straightforward and with lots of statistics,” said Caelean Barnes, whose two-year-old start-up, Gauge, also focuses on companies that want to make their products more noticeable to A.I. agents. “You need to make sure the facts about your business are readily available to agents across all the different surfaces they might look at.” In practice, this means more output: more product descriptions, more blog posts and more social media activity. And sites should be seamless for agents to peruse. New Generation, in partnership with Visa, rates a brand website’s “readiness for interacting with agents” from 1 to 100 by simulating agents’ attempts to navigate it. The score helps brands see what agents can handle and where they get stuck. “Say you’re a bank and agents are asking about your credit card interest rates,” said Jonathan Arena, the New Generation chief executive. “Then a shopper is like, ‘Hey, I want to apply for your top-tier credit card’ but that run failed because their agent couldn’t access the application form. You just lost a customer.” Technical tricksSome strategies are exceptionally technical. Shamny said Limy focuses on how A.I. systems represent language as a series of numbers called vectors. A prompt like “best running shoes for a marathon” gets mapped to a numerical representation of its meaning. A website’s content can also be represented numerically, and Limy helps “identify the vector spaces agents are interested in.” (The start-up also identifies brands’ MVPs — the most valuable prompts — which result in the most purchases). Not everyone is enamored of such wizardry. The models are changing “faster than our ability to try to game them,” said Toby Coulthard, the founder of Jacquard, an A.I. marketing platform. “Maybe these recommendations are going to be true for the next two months but not after that. You have to take it back down to fundamental marketing principles.” Akshat Trivedi, an industry analyst with FTI Consulting, sees parallels to both the mobile commerce transition and search engine optimization bonanza. First movers earned big gains. But eventually, when everyone got there, the market equalized. What is different now, he said, is the speed at which companies are racing to get ahead of what seems to be a new marketing paradigm. “Soon we might see products that launch first on A.I. or discoverability that is driven by brands named in such a way that A.I. picks them up more than others,” said Trivedi. That likely means more marketing budget disappears into the agentic frontier. Ackerman, at the RTA Store, envisions some day soon spending up to $12,000 a month to catch the attention of his bot clientele. “As these systems get more intelligent, we’re going to have no choice but to spend a bit more,” he said.
A soft job report dims odds of an October rate increase. Employers added 29,000 jobs last month, fewer than economists expected, wage growth slowed, and the unemployment rate rose to 4.2 percent. Projections released by the Fed last month suggest most officials see a need for at least one more quarter-point rate increase this year. A.I. leaders agree to police themselves. After a meeting at the White House with President Trump, the leaders of the top A.I. companies agreed to voluntary safety guidelines. The guidelines are similar to those many of the same leaders agreed to three years ago, and drew criticism for not going far enough. Also in Washington: President Trump introduced a government chatbot and plans to name Jay Clayton, the director of national intelligence, as A.I. czar. Anthropic is reportedly planning to go public before Thanksgiving. The company is set to meet prospective investors on Oct. 14, according to Bloomberg. On Monday, Reuters reported that Anthropic’s revenue grew twelvefold to $4.6 billion last year, while its operating loss surpassed $8 billion over the same period, citing a leaked prospectus. More big deals: Oura delayed its I.P.O. The Senate passed a bill to remake college sports. And U.S. bond yields hit their highest level since 2002 on Thursday.
Inside the race to A.G.I.As artificial intelligence agents go rogue and their makers warn of existential risks, an upcoming book by the tech columnist Kevin Roose, “The AGI Chronicles,” explains how we got here. Roose spoke with DealBook’s Sarah Kessler about what executives and policymakers need to understand about the biggest A.I. labs. The conversation has been condensed and edited for clarity. What do you think the average business person misunderstands about the big A.I. companies? The mistake people sometimes make is concluding that these are normal businesses. That the people who run them are interested in maximizing shareholder value or juicing their I.P.O.s. The people making this technology genuinely thought even a decade or more ago that it would transform society, and they were a little scared of that power. So they did weird things. Like what? In the book I report that, when they were setting up Anthropic, Dario Amodei asked his co-founders to take a pledge, a secret pledge, that they would all donate at least 80 percent of their equity to charity. He didn’t want the money and the pursuit of money to corrupt the mission. I also report that Anthropic, for many years, had no plans to commercialize its technology and had explicitly decided not to. They didn’t want to end up like OpenAI, which they had all left because they thought it was straying from the mission. It was only after ChatGPT came out that they finally sort of begrudgingly decided they would make and sell their own chatbot, too. You write when you started reporting the book in February 2025, your sources were evenly split between believing they were building something that would save the world or hurt the world. Has that changed? I think it’s more like 80-20 now on the doom side, and I think partly that’s because the models have continued to get better and more capable. There were a lot of people, even in the A.I. industry, who were hopeful that the models would plateau. I think the Hugging Face attack really freaked a lot of people in the industry out. Lawmakers have been throwing around a lot of potential interventions, even though they’re making no progress toward them. Which of the ideas do you see as most and least feasible? The proposal to put evaluators inside the big A.I. labs is reasonable. I don’t think it’s enough, but I think it is a good starting point. Ultimately, I’m not very optimistic about this administration or this Congress being able to do something on a relevant time scale. But I think the A.I. companies are sincere in their desire to slow down the frontier, and I think we should take them up on it. What about a kill switch? I don’t know that it makes a ton of sense. It’s not like the movies where you can just kind of run over to the server rack and pull out the cord.
DEALBOOK QUIZ Hollywood’s big mash-upThis question comes from a recent article in The Times. Click and answer to see if you’re right. (The link will be free.) Paramount’s acquisition of Warner Bros. Discovery is finally nearing completion, and the combined company is set to debut on Oct. 6. What will it be called? We hope you’ve enjoyed this newsletter, which is made possible through subscriber support. Subscribe to The New York Times. Thanks for reading! We’ll see you Monday. We’d like your feedback. Please email thoughts and suggestions to dealbook@nytimes.com. Follow DealBook on Instagram: @nytdealbook
|
