More Amazon delivery drivers may soon be looking through their next piece of workplace technology rather than down at it.

The company plans to deploy more than 20,000 pairs of smart glasses to its Delivery Service Partners by the end of 2027, giving delivery drivers real-time navigation, package scanning and pet alerts, reducing the need to continually look down at a smartphone.

The glasses represent Amazon’s latest attempt to use AI and automation to remove small inefficiencies from a delivery system operating at enormous scale. But they are also the latest piece of technology being inserted into a journey that increasingly resembles a coordinated exchange between humans, robots and AI, beginning almost as soon as a customer clicks “Buy Now.”

Amazon’s smart glasses use cameras, computer vision and AI-powered sensors to place delivery information directly into a driver’s field of view. Once a driver safely parks, the glasses activate and help identify the correct packages inside the vehicle. They then provide walking directions to the customer’s door, including guidance through more complicated environments such as apartment buildings.

The glasses can also capture proof of delivery and warn drivers about potential hazards. When Amazon’s delivery data indicates a pet may be present at an address, for example, a paw-print icon appears accompanied by an audio alert. The system can also detect when a driver may have left a package at a doorstep that does not correspond with the intended address.

Amazon says more than 500 Delivery Associates have tested the glasses over the past 18 months, completing more than 275,000 deliveries. The company plans to deploy an additional 5,000 devices this year before expanding to more than 20,000 by the end of 2027. Use of the glasses is voluntary for Delivery Service Partners and their drivers.

The device includes a small controller worn in the delivery vest, a replaceable battery designed to last throughout the workday and an emergency button that can connect drivers with emergency services. Prescription and transitional lenses are supported, and a hardware switch provides privacy control.

While the glasses bring AI closer to the customer’s front door, they are only the visible end of a much larger technological system operating behind an Amazon delivery.

Long before the driver reaches the porch, robots may have already moved the product through a fulfillment center. Amazon’s sortable fulfillment centers are typically about 800,000 square feet, large enough to hold millions of individual products. Instead of arranging all of that merchandise in a way that would make sense to a person walking through the building, Amazon digitally tracks where inventory is stored.

When an order reaches a fulfillment center, software can dispatch a robot to retrieve the storage pod containing the item. Hercules, one of Amazon’s workhorse robots, can slide underneath a pod, lift as much as 1,250 pounds and carry the entire shelving unit to an employee.

Hercules does not wander. The robot receives instructions from centralized planning software and uses a 3D camera to identify people, other robots, storage pods and obstacles. Encoded markers spread across the floor help it determine its position and navigate through the facility.

The inventory comes to the worker, and the system identifies the appropriate compartment. The employee removes the item, scans it and sends it along toward packaging.

Robots can handle individual products as well. Amazon’s Vulcan robot combines computer vision with a sense of touch, allowing it to reach into crowded storage compartments and pick or stow products while determining how much force it is applying. Amazon says Vulcan can handle about three-quarters of the different types of items stored in its fulfillment centers.

Other robots perform different jobs. Pegasus moves packages using conveyor belts, while Proteus, Amazon’s first fully autonomous mobile robot, transports heavy carts through areas where employees are also working. Robotic systems using computer vision can sort packages and determine where they should go next.

The scale of the operation continues to grow. In June 2025, Amazon announced that it had deployed its 1 millionth robot, with its robotic network spanning more than 300 facilities worldwide. At the same time, it introduced DeepFleet, a generative AI foundation model designed to coordinate the movement of those robots.

Amazon compares DeepFleet to an intelligent traffic-management system. Rather than allowing thousands of machines to independently choose routes through a crowded facility, the AI analyzes their movements and coordinates more efficient paths. Amazon says the system can improve robot travel efficiency by 10%.

For the customer, much of this remains invisible. A product is selected on a screen, a button is pressed and, sometimes within hours, a package appears outside the door.

Behind that transaction is an elaborate collaboration between people and machines. Robots retrieve shelves, AI coordinates their movements, computer vision helps identify and sort products, workers pick and pack orders, and products are placed in delivery vehicles.

Now Amazon wants some of that AI on the driver’s face.