Amazon's Fulfillment Centers: Generative AI and Operational Efficiency

Amazon leverages generative AI to enhance efficiency, speed, and sustainability in its fulfillment and delivery processes.

### Revolutionizing E-Commerce: Amazon's AI-Driven Fulfillment Centers
* 🏢 Amazon's fulfillment centers efficiently handle packages through a combination of **human labor and advanced robotics**, all within a massive **1,000,000 square foot warehouse** in Northern California.
* 🤖 The integration of generative AI enhances the operations, enabling robots like **Robin arms** and **Pegasus drives** to **optimize the sorting and delivery process**.
* 📦 With a focus on speed, Amazon has transformed delivery expectations from **two-day to same-day service**, leveraging extensive data to improve predictions and delivery routes.
* 🌱 While AI contributes to **operational efficiency and sustainability**, concerns remain regarding potential **job displacement** and **environmental impacts** from data center operations.
* 🚚 Amazon's investment in robotics and AI has led to significant advancements, including the deployment of **750,000 robots** and a network of **390,000 drivers**, enhancing delivery capabilities.
* 📈 Generative AI is pivotal in forecasting demand and optimizing inventory placement, which **minimizes shipping distances** and **improves overall efficiency**.
* 💡 Despite the benefits, challenges such as **workplace safety** and the **environmental costs of AI operations** persist, prompting Amazon to invest in **employee training and safety measures**.
* 🔍 As Amazon continues to innovate with AI, it aims to balance automation with human roles, ensuring that technological advancements translate into **improved service and shareholder value**.

So fulfillment centers ship into this building. This building sorts those packages. Then they flow out to our delivery stations. Inside this 1,000,000 square foot warehouse in Northern California, Amazon packages are handled by hundreds of people and hundreds of robots, all increasingly driven by tech's biggest craze. Generative AI underpins everything we're doing here with perceiving, grasping, and moving products with the Robin arms to managing fleet congestion with our Pegasus robotic drives. For years, Amazon has sped up package delivery - two-day, one-day, and now, more and more, same-day - made possible by more workers but also by rapidly improving algorithms and AI-enabled robots. What you see here is, I like to call it our dance floor. And thanks to being an online retailer from the start, Amazon has a big advantage when it comes to the massive data needed for generative AI. Absolutely. Amazon has been better at it than probably every other retailer out there. Think better predictions of exactly what you're going

to order from where and when. Hundreds of thousands of robots and more efficient delivery routes. But not all the change that could come from generative AI is positive. In the event that they were able to leverage generative AI and fully automate the fulfillment center, I think that would be problematic. The other downside that we don't talk about enough is the negative impact on the environment: the costs of running data centers, the use of electricity, the use of water for cooling. Still, Amazon says AI helps cut costs and its carbon footprint thanks to more efficient planning. It seems subtle, but at this scale, getting like just one more product in the right spot means that it's shipping less distance when you order it. Better speed, lower distance traveled, better sustainability. CNBC visited Amazon's largest California sort center and a same-day warehouse nearby to see firsthand how it's putting AI to work at every step of operations, and sat down with Amazon's head of transportation technology to find out just how far the e-commerce

giant plans to take it. When Amazon Prime launched in 2005, two-day shipping was virtually unheard of. And although it's now standard and free for millions of items if you're a Prime member, it's a grueling logistical lift. Along here, you see robotic arms. They're loading packages with employees onto Pegasus drive units. Those Pegasus drive units are then sorting packages by neighborhood. Steve Armato started at Amazon as a software engineer before Prime in 2001. A lot of the things you see today, those weren't there in 2001. We had five fulfillment centers back then. Now we have hundreds. The delivery vans that you see in the neighborhood, none of that was there. Traditional retailers like Walmart and Target were selling online, but they hadn't started making promises of speed. Back then, you know, mail order when Amazon started, you'd be lucky if you could get something in 2 to 3

weeks. And, you know, Amazon would still promise like a week, and they would yet get it to you in a few days. So that was amazing. So how did Amazon pull it off? The short answer is data. Long before generative AI became all the rage with the release of ChatGPT in 2022, general AI was a huge differentiator for Amazon. As an early online-only retailer, Amazon had a unique ability to collect mass aggregate data on shopping behavior and use it to create algorithms to maximize sales and speedy logistics. We've been working on AI for over 25 years. For employees, a lot of it is around ergonomics and safety. For customers, it's around vast selection, great speeds. Exploiting technology to drive e-commerce sales. That's essentially what Amazon has done since '97. Since the beginning. They are, I would say, hands down the most data-heavy and data-savvy company. It's not that Walmart and Target and Costco and others

don't have their own streams of data, but they're looking at things a little bit differently, and they have much older systems. Amazon is decades younger than its major retail competitors, but its stock value and footprint have grown incredibly fast. Hundreds of warehouses, more than 1.5 million U.S. employees, and more speed. In 2014, Amazon launched Prime Now with some deliveries arriving in an hour or less. Then in 2018, Amazon vastly increased its driver network with the launch of its Delivery Service Partner program, where it contracts driving out to some 4,400 small delivery businesses that employ 390,000 drivers. By 2019, one-day shipping was the norm. Then in 2020, Amazon began using transformer architecture, the backbones of what we know of today as generative AI, to develop models for demand forecasting and supply chain optimization. By 2022, it was rolling AI transformer models into its robotics. All that made shipping times even faster.

Today, drivers are delivering 20 million packages per day across 20 countries, and in the first quarter of 2024, more than 2 billion items arrived the same or next day. 60% of our deliveries for Prime customers in March were same-day or next-day, so there are a lot of those fast orders for our top 60 metropolitan areas. Could you have gotten to that 60% number without generative AI? Well, I think we've been working on this for decades to get to this speed. And it's a combination of engineering, people, processes, and technology. Generative AI is a big unlock for us, particularly for new products where we have sparse or no history for that sales history for that product. It's going to come faster because of generative AI. But all this speed comes at a huge cost, in actual cap ex, but also human labor, a burden that can be reduced, Amazon says, with the use of robots. Before robotics, pickers would need to walk distances

between aisles to pick products, kind of like a library. And now it's being brought to you more like self-service. And so that's great for ergonomics. It's great for less walking. Amazon has faced scrutiny in recent years over its workplace injury record, with federal citations for safety violations and a year-long Senate probe that found Prime Day was a major cause of worker injuries. Amazon has appealed the citations and said the report ignores progress it's made, and it says AI can help. One algorithmic improvement is to take our faster-selling products and place those on the shelves at waist height. That's your ergonomic power zone. So less reaching, less bending. Amazon's big shift to automation started in 2012 with the purchase of Kiva Systems for $775 million. Now, Amazon has deployed at least 750,000 robots, more than double the number it had in 2021. So generative AI helps with prioritization.

So some of the two-day deliveries might stand aside and let the robot with a next-day delivery go on its mission first and take a straight line to its destination. Amazon's next generation of drive units, called Proteus, are fully autonomous. They're actually outside of this fenced area, moving things around. They're using generative AI and computer vision to avoid obstacles and find the right place to stop, the right place to park. And then there's Robin arms, which Amazon says have handled some 2 billion packages so far. Would these 20 Robin arms be able to do what they're doing without generative AI? I think that we would see that they would require a lot more training. And so generative AI has been really a step function improvement in being able to infer from our vast product catalogue about how to handle a given product, even if I haven't seen that product before. And although it's only in a couple of warehouses for now, there's a humanoid robot called Digit that can

grasp and handle items in a similar way to how people can. And Amazon has a new deal with AI startup Covariant, hiring its founders and licensing its models that help robots handle a wider range of physical objects. Of course, this brings up the big question: could Amazon one day replace all warehouse workers if AI helps robots get too capable? There's kind of a balancing act for Amazon. How can they implement automation to improve efficiency and manage labor expenses? But how can they do it in a way that complements their use of humans and doesn't replace them? Amazon says the robots work with people, and they're creating new roles. We're investing over $1.2 billion to upskill more than 300,000 employees by the end of next year. One study found that each robot adopted in manufacturing replaced about three workers. Other research shows that companies that deploy more robots add more jobs overall. So, someone needs to maintain this if it breaks down. Or if something does get dropped on the dance floor,

we have a process and special training to go clean that up. And so, each of those creates new categories of jobs, some of which have higher earnings potential as well. The truth is, at the end of the day, Amazon's responsibility is not to employ, you know, kind of a million Americans, even though it does. Their responsibility is to their shareholders. Another big way to please shareholders is to cut down on the huge amount of time and money it takes to get inventory from sellers to customers. To do this, Amazon has always used algorithms to predict how much of what inventory is needed, when and where. Every product has like a regional nuance. We recently regionalized our entire national network. And by doing a regional network, that means that products are more likely to ship from fulfillment centers close to you. What's new with generative AI is the ability to predict where to place brand new items. So, we're able to use generative AI to create a link between products we have seen before, where we do have

a sales history, and a new product we haven't seen before yet, and get it in the right place the first time. So when we place a product in the right place ahead of time before you click buy, it's traveling less distance, which is a win for speed and sustainability. Amazon also says AI is helping sustainability with a specific model that makes better choices about which packaging to use, and by reducing the number of damaged items that get sent to and returned by customers. Amazon says its AI is three times better than humans at identifying damaged products. We ship billions of packages. We have the data about those packages. So we're able to use computer vision together with generative AI and that vast product and package data, to then detect damages and being able to sideline a package if we think it might be damaged before we ship it to a customer. But training and running AI is itself a carbon-intensive process, a fact that could make it hard for Amazon to achieve its 2019 climate pledge to reach net

zero carbon by 2040. By 2027, AI servers are projected to use up as much power every year as a small country. And Amazon Web Services has data centers filled with servers running AI workloads, although this also gives it an edge over other e-commerce players because it can train its AI in-house. Amazon has also invested $4 billion in AI startup Anthropic, which makes chatbot Claude a competitor to OpenAI's ChatGPT. And Amazon makes its own AI-focused microchips and its own generative AI tools for developers, which are used in operations. We use tools like Amazon Q, Amazon Bedrock. That allows us to evaluate different models against the, you know, what does good look like? So what are the metrics for success of this business application? One big metric for success that shareholders are watching: if Amazon's huge investment in AI will translate to profits. I have yet to see huge lift in anybody's retail business due to generative AI, including Amazon.

I think that a lot of their biggest impact has happened because of the earlier investments, not necessarily some of these more recent investments. I've seen a lot of hype, but no actual numbers. One area Amazon is hoping AI will translate to true savings is the most expensive part of the delivery process: getting that package the last mile to your door. Amazon is now using more than 20 machine learning models to figure out the most efficient routes for its delivery drivers. If there is more congestion on a road or if a road is closed, AI is able to help us determine whether that diversion is still there or take that different route. In 2021, CNBC talked to Amazon drivers about the pressures and pitfalls of the job, from dog bites to urinating in bottles to save time. People are running through stop signs, running through yellow lights. Everybody I knew was buckling their seat belt behind their backs because the time it took just to buckle your seat belt, unbuckle your seat belt.

every time was enough time to get you behind schedule. The hope is, with better routes and vehicle coordination, drivers will feel less pressure to cut corners or skip breaks. It should enhance the driver experience, but it will still be challenging. In 2022, Amazon also rolled out new fully electric Rivian vans and now has 15,000 of them across the U.S. They're equipped with large screens where the new mapping and routing is displayed, as well as AI-enabled cameras that watch the road, sides of the vehicle, and the driver. There's no camera recording if the driver is not driving, and there's a privacy mode. Another area where privacy often comes up is around the huge amount of data it collects on shopping behavior. Now, with generative AI, that data can be used to generate better hyper-personalized product recommendations to a shopper, thanks to new developer tools like Amazon Personalize. What do you say to people when they're concerned about

their privacy? Or that it feels creepy that Amazon predicts their shopping behavior so closely? Well, we're looking at aggregate sales history. We're looking at geographic regions and their behavior. If a particular customer were in San Francisco, it's not about that one customer. It's about what the aggregate behavior in that area is. Sellers can also use gen AI to write more targeted product listings, or to generate images of their products in different seasonal and lifestyle settings. For shoppers, last year Amazon.com started populating with AI-generated review highlights. Any product that you look at is going to have reviews, hundreds or even thousands. What I love about AI review summarization is it gives you just a couple of quick bullet points, and so that's helping me make a more informed buying decision. And in February, Amazon launched a new gen AI-powered conversational shopping assistant called Rufus, to further streamline those product recommendations. Some consumers may be put off by product recommendations that are based on their purchase

history. Now Amazon may have to, if they don't already implement, an opt-out, some sort of feature where a consumer could say, you know, please don't look at my purchase history when giving me recommendations. Despite concerns, Amazon is committed to injecting generative AI into every possible step of operations to boost speed, efficiency, and eventually, it hopes, its bottom line. If you can reduce packaging, is that a good thing? Yes. If you can reduce the amount of time and the amount of miles you have to travel to get from point A to point B, is that a good thing? Yes. These are all good ideas, you know, kind of definitely feel free to show us the receipts at any time.

How Amazon Is Delivering Packages Faster With The Help Of Generative AI

Welcome to the official YouTube channel for CNBC. From 'Wall Street' to 'Main Street' to award winning original documentaries and Reality TV series, CNBC has you covered. Experience special sneak peeks of your favorite shows, exclusive video and more.

For real-time news updates from around the world, visit: http://www.cnbc.com

For CNBC's primetime TV lineup, visit http://tv.cnbc.com

For CNBC's International YouTube channel, visit: http://www.youtube.com/cnbcinternational

CNBC

Amazon's Northern California warehouse, spanning 1,000,000 square feet, uses a combination of human workers and robots, driven by generative AI, to sort and deliver packages.

The company's rapid package delivery has evolved from two-day to same-day service, leveraging data and AI for better predictions and efficient routing.

While generative AI has its advantages, such as cost-cutting and sustainability, there are concerns about its potential negative impact, including job automation and environmental costs.

Amazon's Success Through Data and Technology

Amazon's rapid delivery success is driven by its unique ability to collect and analyze shopping behavior data.

The company's growth has been fueled by a vast network of warehouses, a large driver workforce, and the use of AI for demand forecasting and supply chain optimization.

Today, Amazon delivers 20 million packages daily, with a significant portion arriving the same or next day, and is leveraging generative AI to further enhance its speed and efficiency.

Amazon's Shift to Automation and AI in Warehousing

Amazon is transitioning to a self-service model in its warehouses, aiming to improve ergonomics and reduce walking for workers.

The company has faced scrutiny for its workplace injury record, leading to a significant increase in robot deployment since 2012.

Amazon's use of generative AI is enhancing automation, prioritizing tasks, and enabling robots to handle a wider range of physical objects, while also investing in upskilling employees and creating new roles.

Amazon's Use of AI for Efficiency and Sustainability

Amazon's algorithms and regionalized network aim to streamline inventory distribution, while generative AI helps predict the placement of new products.

The company's AI also plays a role in sustainability by optimizing packaging choices and reducing damaged items.

However, the carbon-intensive nature of AI training poses a challenge for Amazon's climate pledge, and the company's significant investment in AI raises questions about its potential to drive profits.

Amazon's AI-Driven Delivery Transformation

Amazon's early investments have had a significant impact, with AI now playing a crucial role in optimizing delivery routes for drivers.

The company is using over 20 machine learning models to navigate congestion and road closures, aiming to reduce pressure on drivers.

Additionally, Amazon has rolled out 15,000 fully electric Rivian vans with AI-enabled features, while also using generative AI to provide hyper-personalized product recommendations for shoppers.

Amazon's Use of Generative AI for Shopping

Amazon analyzes aggregate sales history and geographic behavior to predict shopping patterns, rather than focusing on individual customers.

The company uses generative AI for targeted product listings, image generation, and AI-generated review highlights to assist shoppers in making informed decisions.

Amazon's conversational shopping assistant, Rufus, was launched to streamline product recommendations, but the company may need to implement an opt-out feature for customers concerned about their purchase history being used for recommendations.