Amazon is moving a wearable from the pilot phase into the network. The company plans to deploy 5,000 smart delivery glasses this year and expects more than 20,000 in the field by the end of 2027, according to an update tied to its Delivery Service Partner program and reported by Supply Chain Dive on November 4. More than 300 delivery service partners and hundreds of their drivers have used early units since 2025, and Amazon says the hardware will be provided at no cost, with participation optional for both contractors and drivers.
How the device works
The glasses are built around a narrow problem: a driver carrying packages has no free hand for a handheld device, and every glance down at a screen is a moment not spent watching the street. A small display inside the lens activates automatically when the van is parked, showing how many packages belong to that stop and confirming each one as it is scanned by glancing at the label barcode. Walking directions and delivery instructions, including gate codes, access notes and preferred drop-off spots, appear in the field of view rather than on a phone. Proof-of-delivery photos are captured by looking at the package and pressing a button on a chest-worn controller.
The hardware decisions are pragmatic rather than fashionable. Computing and battery live in a small controller worn in the delivery vest, which keeps weight off the frame across an eight-hour shift, and the battery is swappable so a route does not end early. The controller carries a dedicated emergency button. The glasses support prescription and transition lenses, and later versions add an LED headlamp, hands-free scanning and proactive address verification. Amazon says the display blanks whenever the vehicle is moving, a design choice aimed at both safety and regulation.
Privacy, by design and by argument
Because the wearer is a worker walking through residential streets, Amazon has built the privacy case into the product. The camera captures still images only, with no video and no audio. Faces and license plates are detected and blurred before storage, imagery is encrypted and restricted to a limited group, retained briefly and then deleted, and no third party processes it. Drivers can switch the camera off with a hardware privacy switch at any time, including on breaks, and the glasses do not capture while the vehicle is in motion. Amazon also says drivers who decline to wear them face no penalty.
The arguments that will not be settled by a spec sheet are harder. Driver advocacy groups have spent years documenting how Amazon-adjacent technology, from in-van cameras to handheld scanners, functions as monitoring infrastructure regardless of its stated safety purpose, and a wearable camera worn by a contracted worker extends that debate into new territory. Those are open questions that will shape how the rollout is received by regulators.
The economics underneath
The business case is not about the glasses. Amazon has said its transportation vice president estimated savings of up to 30 minutes across an eight-to-ten-hour shift by eliminating repetitive device handling, a figure the company itself describes as preliminary and drawn from initial testing at limited scale. Spread across scores of stops per route and millions of routes per quarter, even a fraction of that compounds into labor hours, and labor is the dominant cost in a network where the last mile is performed by contractors operating under Amazon pricing.
That is the strategic frame analysts have fastened onto. Amazon fulfillment and shipping costs totaled roughly $211 billion in 2025, split almost evenly between the two, against a North America operating margin near 7 percent, while its cloud business earns margins several times higher on a fraction of revenue. Efficiency hardware aimed at the hundred yards between the van and the door is one of the few levers that improves that ratio without adding volume, alongside warehouse robotics and routing software. Amazon has also invested heavily in the mapping layer that makes the glasses useful: its generative mapping system has cataloged more than 200 million parking locations, 2.8 million building entrances and 85,000 mailrooms and lockers since 2024.
The contractor question
The deployment lands on a structurally awkward part of the network. Amazon does not employ the drivers; it contracts with delivery service partners, nominally independent small businesses that hire, insure and manage their own teams. Supplying hardware, training and software directly to those drivers deepens Amazon involvement with a workforce it says it does not control, a tension that has produced litigation over contractor autonomy for years. The counter-argument is that Amazon is spending real money on the program: an additional $1.9 billion earmarked for 2027 brings the eight-year total to $21.7 billion, with driver pay reported near $24 an hour nationally and serious crashes down more than 23 percent last year.
Whether 20,000 units materialize depends less on the hardware than on three operational variables: whether the time savings hold at network scale rather than in demos, whether battery and weight issues stay solved across full shifts in bad weather, and whether state rules treat head-worn displays as distractions. Illinois has already moved to classify AI smart glasses as banned devices while driving, with no hands-free exemption, a reminder that a product designed to keep eyes up can still be regulated as a screen.
What it means for logistics operators
For third-party logistics providers and last-mile contractors outside Amazon, the signal is that enterprise wearables have crossed from demonstration to procurement. The metrics that will decide adoption are unglamorous: dwell time at the door, wrong-address rates, proof-of-delivery disputes and seconds per stop. Operators evaluating similar hardware should insist on measured baselines before rollout, because vendor estimates of time saved are usually generated under favorable conditions and rarely survive a rainy Tuesday in a dense apartment complex. The more durable lesson is architectural: the glasses only work because a decade of geospatial mapping sits behind them, and any operator without that layer will find the headset is the cheap part.
