How Amazon Robotics Changed the Landscape of Fulfillment

Amazon changed the industry baseline for fulfillment. Fast delivery, broad product selection, and easy returns became expected, pushing retailers to rethink what their warehouses could do. This is widely referred to as the “Amazon Effect”.

Amazon’s robotics are a big part of how that story came to life. Its fulfillment operations combine mobile robots, robotic arms, and software that moves inventory, prioritizes work, and supports faster order execution. Together, those robots and software help Amazon absorb change, whether it’s from new products, shifting order priorities, or the volume spikes that come with Black Friday and the holidays.

Amazon continues to build new capabilities for its own warehouses and its customers outside of just robotic hardware. It is extending those competencies in two ways: offering fulfillment services to outside companies and using AI to improve the software behind its automation.

Key Takeaways:

• Amazon runs experiments at scale: Amazon has deployed more than a million robots, from Kiva AMRs to robotic arms. However, breadth comes with trial and error, as seen when it pulled Blue Jay off the warehouse floor.
• AI is helping coordinate Amazon’s robot fleet: Amazon is deploying agentic AI to help teams make decisions and direct its robots. Tools like DeepFleet are designed to improve robot travel efficiency, while Project Eluna helps warehouse teams surface system slowdowns and evaluate solutions.
• Amazon sells fulfillment now: Amazon Supply Chain Services gives brands another fulfillment model to weigh against building or buying their own automation.

Want to learn more? Keep reading.

The History and Growth of Amazon Robotics

Amazon’s journey into robotics began with its 2012 acquisition of Kiva Systems for $775 million, marking a turning point in warehouse automation. Kiva robots, classified as Autonomous Mobile Robots (AMRs), were designed to zip across warehouse floors, transporting lightweight shelves with ordered goods to human operators for picking. This system significantly boosts efficiency by reducing manual picking time and improving order accuracy. It also creates a safer, more ergonomic work environment by minimizing the need for employees to bend and lift heavy items or work near hazardous equipment like forklifts.

Amazon has continued to invest in warehouse automation, and its warehouses run everything from mobile robots and robotic arms to AI software that coordinates fulfillment. In fact, Amazon now runs more than one million robots, a milestone it passed in 2025.

One key robotic automation system Amazon introduced across its global warehouse network after acquiring Kiva Systems was its Sequoia system. Amazon’s Sequoia system combines mobile robots, gantry systems, robotic arms, and an ergonomic workstation to store inventory in totes and bring it to employees for picking.

Amazon says Sequoia can identify and store newly received inventory up to 75% faster and reduce the time needed to process an order through a fulfillment center by up to 25%.

The company also has various robotic arms. One of note is Sparrow, their first robot that allowed them to handle individual products rather than full totes, using computer vision and AI to detect, select, and move items from inventory bins.

Vulcan, another robotic arm built by Amazon, incorporates force feedback similarly to how a human hand might to avoid damaging items.

Picking is usually a challenge to automate because mimicking the human grasp is hard. A human hand adjusts its pressure as it works, gripping an item tightly enough to move it without damaging it, and that reactive control has been difficult for robots to reproduce. 

The way Vulcan approaches picking is by using a camera to determine which item to pick and where to grab it, and it simulates a sense of touch via sensors so it can judge how much force to use. Paddles then direct items onto its conveyor belt, where they eventually fall into a bin. Because it picks this way rather than with a hand, Vulcan can only handle about 75% of inventory items, and it notifies its human peers when it needs assistance.

Amazon continues to advance its robotic software as well.

In 2025, Amazon introduced DeepFleet, a generative AI foundation model designed to coordinate robot movement across its fulfillment network. Amazon says the model is intended to improve robot travel efficiency by 10%.

Amazon developed a similar agentic AI assistant in 202. Project Eluna is an agentic AI assistant that pinpoints system hold-ups, determines resolution paths, and communicates this information to teams on the warehouse floor.

Beyond robotics and software, Amazon is also looking for ways to capitalize off their own advanced network.

This year, Amazon announced it is extending its fulfillment capabilities through Amazon Supply Chain Services. That gives brands another model to weigh as automation becomes part of the service offering. (We looked at what that means for 3PLs in this article).

The Ripple Effect of Amazon Robotics

How the Kiva Acquisition Changed the Automation Market

When Amazon adopted Kiva Systems, it left a gap in the market for businesses seeking advanced robotic solutions. 

Amazon’s competitors needed technology to gain some edge back. Other organizations were inspired to adopt an automated storage and retrieval system (AS/RS) as well.

Why Flexibility Became a Priority in Warehouse Automation

Companies soon found that traditional AS/RS could be difficult to adapt. Scaling the operation or changing channel strategies could require complex and costly changes.

Exotec saw the need in the market for a flexible AS/RS, and stepped up. We developed the Exotec Skypod System. Our Skypod robots move goods between racks up to 45 feet high and ergonomic workstations, while the modular design supports expansion without rebuilding the whole system. It functions as the core of our end-to-end warehouse automation solutions.

Companies are also looking for flexible ways to approach an automation investment. Robotics-as-a-service (RaaS) is one market model in which companies access robotic systems without committing the full upfront capital required for a warehouse transformation. 

However, while RaaS can reduce the upfront capital commitment, it does not remove the work of preparing the site for automation. A facility may still need floor remediation for robot travel, along with integration between the robotic system and the software and processes that manage the warehouse. 

Companies looking into RaaS need confidence that the provider can handle floor remediation, deployment, integration, and issue resolution, and will stay accountable when problems arise.

How Warehouse Automation Is Moving Toward AI

Amazon’s investment in AI has made AI harder for the warehouse industry to ignore.

A 2025 Mecalux and MIT Intelligent Logistics Systems Lab survey of more than 2,000 logistics professionals across 21 countries reported that 60% of warehouses already use AI within their warehouse. 92% said they were implementing or planning new AI projects.

However, while Amazon can afford to test and iterate AI in the early stages, other high-volume warehouses may have less room for disruption.

AI functions best when small mistakes for the sake of learning are tolerable, and when a short outage can cost millions in overtime, expedited shipping, contract penalties, and canceled orders, most warehouses can’t take that risk.

What Amazon Robotics Can Teach About In-House Automation Risk

Amazon’s own record with new technology over the last two years shows that not every in-house technology lasts.

In 2025, Amazon introduced Blue Jay, a ceiling-mounted system with multiple robotic arms that picked, stowed, and consolidated items in a single workspace. 

It was pulled from operations within months, reportedly because it cost too much and proved too complex to build and deploy. 

Despite the machine getting sunsetted, parts of Blue Jay continue to be used within the Amazon robotics ecosystem. Amazon kept the underlying technology and folded it into newer systems that are simpler to deploy and scale, including Flex Cell and Orbital, a warehouse system that is more easily adaptable.

For Amazon, this is the cost of innovation. Other organizations may not have the same capital or operating margin to develop, test, and retire complex systems inside live operations.

Overall, the retirement of Blue Jay tells us that demand for faster fulfillment will keep pushing robotics forward, but the systems built to adapt will outlast the ones built fastest.

What Fulfillment Automation Changes for Warehouse Operations

Why invest in fulfillment automation? The gains show up across the warehouse: people can spend more time on higher-value work, warehouse work can be less physically demanding, and orders can move faster with fewer costly errors.

Goods-To-Person Robotics Upskills Human Roles 

Amazon Robotics has also raised questions about the future of what work will look like for warehouse workers as the adoption of robotics and AI continues to grow. 

Robots are best at the parts of warehouse work that are repetitive, slow, or unsafe. Picking and sorting are most of that work, which is why many organizations choose to automate those processes first.

In large facilities, some pickers can walk multiple miles a day. That physical toll can make warehouse roles harder to fill and retain. Goods-to-person robotics removes the walking and puts people in jobs that are less strenuous, reward judgment rather than endurance, and ultimately result in less turnover.

 In our 2025 Warehouse Workers Sentiment Report, based on a survey of over 400 US warehouse floor workers,  63% reported higher job satisfaction working with automation than in manual operations. 

A better workplace experience is only half of it. Automation also frees people to move to more valuable parts of the warehouse. 

One clear example of this is when Ariat, a performance footwear and apparel brand, moved 80% of its picking labor to higher-value work after adopting an Exotec robotic solution, while also raising picking productivity by 10x.

The roles automation tends to create include:

• Process improvement: people who find and remove waste in how the operation runs.
• Data analysis: automation produces a lot of data, and someone has to turn it into decisions. 
• Technical training: robots need people who can run and maintain them.
• Order customization: demand keeps growing for personalized goods and packaging.
• Quality control: someone has to catch defects and keep the operation compliant.

This type of higher-value work often leads to more pay. 49% of the warehouse workers that we surveyed received a raise because they worked with automation, which counters the assumption that automation threatens job stability.

Warehouse Automation Improves Safety and Ergonomics

Repetitive lifting causes injuries over time. When robots do the lifting, workers face less physical strain and injury rates fall. Fewer people walking the floor also means fewer collisions between workers and equipment.

Our 2025 Warehouse Workers Sentiment Report found that three in five respondents who are working with automation reported fewer workplace injuries, and 59% reported less physical strain.

Economic Impacts: Productivity, Operational Costs, and Efficiency

The economic impact of robots in fulfillment centers is clear. 

Unlike humans, robots can run through multiple shifts without needing a break, resulting in improved throughput and the time it takes a customer to receive an order.

The use of robotics also enables efficient space utilization. Companies want facilities close to customers, and urban real estate is expensive. Robots work in narrower aisles and reach higher shelves than people can, so the same inventory fits in a smaller building.

Accuracy affects the economics too. A 1% error rate across one million orders produces 10,000 errors. Those errors can drive rework, returns, replacement shipments, customer-service costs, and erosion of customer trust. To evaluate fulfillment automation ROI, measure how much automation reduces the error rate and the cost of the errors that remain.

Amazon’s advantage comes from scale: it can test ideas in live operations, absorb failed bets, and carry useful technology into the next system. 

Most do not have Amazon’s margin for error. When an automation project fails, capacity is delayed, capital is tied up, and customers may receive orders late or not at all. 

Their automation partner can either add to that risk or help reduce it.

Exotec is an OEM systems integrator that uses standardized, proven warehouse solutions to reduce implementation risk for customers. As the developer of the Skypod system, we know how to best integrate our own technology into an end-to-end warehouse solution and deliver optimal throughput results. 

To see what that Exotec solution looks like in practice, take our virtual warehouse tour.