How Amazon is using generative AI to drive more same-day deliveries

Over the years, Amazon Setting the standard for package delivery. When Prime launched in 2005, two-day shipping was unheard of. By 2019, one-day delivery had become the standard for millions of items. Now, the retail giant is turning to generated artificial intelligence to drive more same-day shipments.

Amazon is using the technology to optimize delivery routes, build smarter warehouse robots and create a more ergonomic environment for workers, said Steve Armato, Amazon's vice president of shipping technology and services. , and better predict where to stock new items.

During an exclusive tour of Amazon's largest California sorting center in Tracy, Armato told CNBC that 60% of Prime orders in March were delivered the same day or next day in the top 60 U.S. metropolitan areas.

“It may seem subtle, but at this scale, just having one more product in the right location means less distance to ship when ordering,” Amato said in an interview at the warehouse.

In 2020, Amazon began developing demand forecasting and supply chain optimization models using the Transformer architecture, which is the backbone of generative artificial intelligence as we know it today.

“Generative artificial intelligence is the next big change in technology,” Amato said. “It’s pretty remarkable and we’re already applying it in a very practical way across our operations.”

But not all the changes generative AI might bring to the e-commerce giant are positive. Analysts told CNBC there are concerns about the high-energy demands of generative artificial intelligence and its ability to allow robots to replace Amazon's workforce.

While these robots use a series of QR codes to navigate, Amazon's next-generation drive unit, called Proteus, Completely autonomousthe company said.

“They are using generative artificial intelligence and computer vision to avoid obstacles and find suitable parking spots,” Amato said.

As part of the company’s artificial intelligence strategy, Amazon implemented trade Partnering with artificial intelligence startup Covariant. Amazon hired the startup’s founders and licensed its models to help robots handle a wider range of physical objects. Amazon is also developing a bipedal robot called Digit that can grab and handle items in a human-like manner.

CNBC saw an array of 20 robotic “Robin” arms that use computer vision to determine how much pressure to use when picking up packages of various shapes and sizes. Amazon says generative AI teaches the arms how to handle products they've never seen before, based on data from similar products in Amazon's vast catalog.

Similar models are used to better assess damaged items and prevent them from being shipped out. Amazon’s artificial intelligence is three times better The company says it can identify damaged products better than humans.

Tom Forte, senior equity analyst at Maxim Group, said introducing more robotics with generative artificial intelligence without replacing human workers would be a balancing act for Amazon.

“How do they implement automation to increase efficiency and manage labor spend, but how do they do it in a way that complements human use rather than replaces humans?” Forte said.

Amato said robots are not replacing workers but easing their burdens and creating new roles. Amazon said it plans to spend $1.2 billion to upskill more than 300,000 workers by the end of 2025 as generative artificial intelligence and robotics transform the company's processes.

“If (the robot) breaks down, someone needs to maintain it,” Amato said. “Or if something does get dropped on the dance floor, we have a process and special training to clean it up. So each of these creates new job categories, some with higher earning potential.”

Amazon has faced scrutiny in recent years for its injury record, with the federal government cited for safety violations and a yearlong Senate investigation finding that Amazon's annual Prime Day sale was the “leading cause” of worker injuries. Amazon appealed the charges and said the report ignored the progress it had made.

Amato said many of Amazon's robots move tall boxes of items to workstations where employees pick and pack them, reducing the amount of time humans spend walking. Artificial intelligence also reduces the need for workers to reach and bend, he said.

“One algorithmic improvement was to place our faster-selling products at waist height on the shelves,” Amato said. “That's your ergonomic strength zone.”

With all those robots and workers, Amazon delivers more than 2 billion items According to the company, it will take place on the same day or the next day in the first quarter of 2024.

Amazon has been using algorithms to predict how much inventory it will need, when and where it will be needed. The company said it is using generative artificial intelligence to predict the best placement for previously unsold items.

“When we place a product in the right place ahead of time before clicking to buy, the distance it travels is reduced, which is a win for speed and sustainability,” Armato said.

Amazon Web Services has data centers full of servers running artificial intelligence workloads, giving the company an advantage over its retail rivals because it can train its artificial intelligence in-house. As an early online retailer, Amazon is a leader in collecting vast amounts of aggregated data on shopping behavior and delivery logistics. Amazon is now using this data to create artificial intelligence models for use in everything from supply chain optimization to warehouse robots, according to the company.

“Not that Walmart and Target and costco And other people don't have a lot of data of their own, but they look at things a little differently and their systems are much older.

It’s unclear how environmentally friendly generative AI will be in the long term. This is because training and running generative AI is a carbon-intensive process, and by 2027, the world’s AI servers are expected to use Annual electricity production is comparable to Sweden or the Netherlands.

This conflicts with Amazon's 2019 pledge to achieve net zero carbon by 2040.

The company claims that the use of artificial intelligence helps reduce the carbon footprint of parcel delivery. Amazon is reducing carbon emissions by using more than 20 machine learning models to improve mapping of its vast network. 390,000 delivery drivers, The company said it anticipates road closures and chooses more efficient routes.

In addition to warehouses, Amazon has also introduced generative artificial intelligence to help sellers and shoppers.

The company's new Amazon Personalize AI tool generates hyper-personalized product recommendations. Sellers can also use generative AI to write highly targeted product descriptions or generate Product pictures under different “season and lifestyle” environments.

For shoppers, Amazon will start populating its site in 2023 AI-generated product review summariesin February, the company launched a conversational shopping assistant powered by generative artificial intelligence called Rufus.

In addition, Amazon said it also Investment of US$4 billion AI startup Anthropic has developed the chatbot Claude as a competitor to OpenAI's ChatGPT. Amazon also produces its own artificial intelligence-focused microchips and its Your own generative AI tool For developers, it is also used operationally, the company said.

Whether Amazon's massive investment in generating artificial intelligence will translate into profits remains an open question.

“I haven’t seen anyone’s retail business get a huge lift from generated artificial intelligence, including Amazon,” Kodali said. “I think their biggest impact will be largely due to early investment, and Not necessarily some recent investment.”

Watch the video to learn more about how Amazon uses artificial intelligence.