Amazon confirmed Wednesday that it laid off an unspecified number of employees in its artificial general intelligence (AGI) organization, the division working on the company’s advanced AI models.

The move, first reported by Reuters, comes as the company invests heavily in programs to help businesses implement AI effectively, including a $1 billion initiative to embed AWS engineers with customers building agentic AI systems.

It’s part of a larger shift in the industry as tech giants and AI frontier labs look to make sure the enormous sums they’re spending on AI pay off in tools businesses actually use.

In a statement, an Amazon spokesperson said building large AI models remains “one of the most important things we’re working on,” but said the company is also “sharpening our focus on the initiatives that matter most for customers, so we can move faster on what counts.”

“That focus means some difficult decisions, including eliminating some roles within parts of our AGI organization, even as we continue to invest in the areas most important to our customers’ future,” the spokesperson said.

It’s the latest in a series of changes in Amazon’s AGI group, which despite its name has always been focused more on frontier models than on what the industry considers AGI, the still-theoretical systems that would match or surpass human intelligence.

Rohit Prasad, the senior executive who oversaw Amazon’s AGI work, left the company late last year, and AGI Lab head David Luan departed in February. In December, Amazon folded the AGI group into a larger organization led by senior vice president Peter DeSantis that also includes chip development and quantum computing.

The cuts are the latest in a series of smaller reductions since January, when Amazon eliminated 16,000 jobs across the company. Amazon said U.S. employees whose jobs are cut will receive 90 days of pay and benefits, outplacement support and transitional health coverage, along with eligibility for severance.

[Update, Friday, July 17: Brown is joining Meta, the WSJ reported. More here.]

Dave Brown, who joined Amazon Web Services as one of its earliest EC2 engineers and rose to lead its compute, AI and machine learning services, is leaving after nearly 19 years. 

AWS CEO Matt Garman told employees in a memo posted publicly Wednesday that Brown will depart at the end of July for an unspecified “new role outside of the company.” Amazon exec Dave Treadwell, who joined the company in 2016 after 27 years at Microsoft, will take over the group Aug. 1.

Brown’s exit comes about three months after Amazon promoted him to senior vice president. Brown had been on the company’s senior leadership team since 2023.

His tenure stretched back to the early days of the cloud. He joined AWS in 2007 in Cape Town, South Africa, where Amazon based part of its early EC2 engineering, before relocating to the Seattle area.

In an interview with GeekWire earlier this year, as the company marked the AWS 20th anniversary, Brown recalled Amazon CEO Andy Jassy, then the company’s top cloud executive, gathering the small Cape Town team in those days and telling them the business could one day be worth a billion dollars.

Brown said he could barely grasp the figure at a time when the service was bringing in tens of dollars a day: “I couldn’t even imagine how much a billion dollars was. It sounded like a lot of money.”

AWS today runs at roughly $150 billion in annualized revenue, and grew 28% in its most recent quarter — its fastest pace in nearly four years.

Brown’s role grew with the business. After starting as an engineer on EC2, or Elastic Compute Cloud, he went on to lead its broader compute organization, including close collaborations with the executives running Amazon’s custom silicon business. His purview also expanded to include the machine learning and AI services now central to AWS, such as the Bedrock and SageMaker platforms.

Treadwell has run Amazon’s eCommerce Foundation, the technical backbone of the company’s online retail operations, since joining in 2016. Before that he spent 27 years at Microsoft, where as a corporate vice president he worked on Windows, Xbox, and the .NET software framework.

In his memo, Garman described Treadwell — known internally as “Tread” — as one of AWS’s largest and most vocal internal customers, someone who pushed the cloud group to innovate and will now lead it.

Brown will remain through the end of July to help with the transition. In his own farewell note, he said it felt like the right time to begin a new chapter. “I’ll be cheering you all on from the sidelines,” he wrote.

U.S. shoppers spent a record $26.4 billion across all retail sites during Amazon’s four-day Prime Day event, and for the first time, the people most likely to complete a purchase were those who arrived from AI chatbots.

It’s the latest twist in a high-stakes bet by Amazon. The AI assistants now sending retailers their best-converting customers are the same ones Amazon has worked to keep away from its own store, hoping to keep shoppers coming directly to Amazon.com and using its own on-site AI assistant instead.

Adobe reported over that weekend that visitors who clicked through to shopping sites from AI assistants were 40% more likely to make a purchase during the four-day event than those showing up through search, email or social media.

AI still accounts for a small fraction of total shopping traffic, but a trend is starting to emerge. In the past, shoppers sent by AI were the least likely to buy, according to Adobe’s data. The change suggests that ChatGPT, Claude, Gemini and others are becoming more effective at giving shoppers the information they need to buy with confidence.

Those figures span all of U.S. retail — “Prime Day” has become much more than a day, and much bigger than Amazon alone. The distinction matters, because Amazon has taken a different path than many of its rivals. While Walmart, Target and others have opened their catalogs to outside AI assistants, Amazon has kept them out.

Agentic AI drives less than 1% of traffic across every major online store, but Amazon’s share is the lowest of the group, at about 0.4%, according to J.P. Morgan data.

That’s by design: Amazon sued Perplexity, for example, over its browser that shopped on customers’ behalf, and won a preliminary injunction barring the tool from the logged-in parts of its site, arguing that unauthorized shopping agents degrade a trusted experience. Perplexity is appealing.

Amazon has separately blocked ChatGPT’s crawlers from reading its listings — even as it has begun buying ads inside ChatGPT to bring shoppers back, a move first spotted by Marketplace Pulse founder Juozas Kaziukėnas and reported by Business Insider and Modern Retail.

On Amazon’s most recent earnings call, in April, CEO Andy Jassy said the company was in talks with the AI companies to come up with a better experience between Amazon and third-party agents to “find something that works for customers and all the companies.”

In the meantime, Amazon is focusing on its own AI assistant.

The tool — launched as Rufus and folded in May into a service called Alexa for Shopping — has drawn more than 250 million users, with monthly users up more than 115% over the past year, the company said. Customers who use it while shopping are more than 60% more likely to buy, and Amazon Web Services has said the tool drove nearly $12 billion in incremental sales last year.

Jassy said on the earnings call that third-party agents weren’t good enough yet — that they lacked a shopper’s history and often couldn’t get prices right — and that people would gravitate to whichever assistant knew them best. That’s the opening Amazon is going after with its own AI chatbot and related tools on Amazon.com.

“We are aiming to have it be the best shopping assistant anywhere,” Jassy said.

The strategy reflects one of the ways Amazon is increasingly making money. Advertising is now among its most profitable businesses. J.P. Morgan expects it to bring in about $83 billion in revenue this year and, because the margins are high, to account for roughly a third of the company’s operating income.

That advertising revenue depends on Amazon getting shoppers to browse its own site rather than handing the decision to an outside chatbot it doesn’t control.

The big question long-term is whether Amazon can maintain its own role as a primary destination for shoppers and avoid becoming just another selection on a chatbot’s shelf.