Arrive AI installs two office delivery endpoints

Arrive AI says it has installed two AP3+ delivery endpoints at a commercial office building in Central Indiana, its first office installation. Announced on 29 October, the units sit near the employee entrance. The company plans to learn from daily use by workers and building operators.[1]

The product combines secure storage with computer vision running locally on an NVIDIA Jetson Orin Nano. Arrive says it detects a package’s presence, orientation and status. Its software also handles events such as whether the door is open and the state of a robot.[2]

The handoff can happen at two different times: a courier leaves the goods, then a recipient or compatible machine collects them later. The office announcement supplies no measured delivery-time results.[1]

What a building changes

An older University of Washington study shows why the receiving point deserves attention. In spring 2018, researchers tested a shared parcel locker at Seattle Municipal Tower, with ten business days of observations before installation and ten afterwards. Thirty-six tenants volunteered; UPS and USPS participated.[3]

The report’s 78% time reduction needs its denominator: the comparison was a single locker delivery of one parcel against eight door-to-door deliveries. The authors describe a best-case demonstration, rather than an expected saving for every delivery. It was a test of a collection point, not of Arrive’s AI.[3]

Placement mattered. The intended location beside the loading bay was rejected because it did not meet accessibility requirements. The locker went by another entrance, leaving drivers a journey between floors. The research also recommended rules for how long parcels could occupy scarce locker space.[3]

For the Indiana installation, the useful questions are practical: does collection become easier, how long do goods remain inside, and when does package recognition prevent a failed handoff? Each measures a different contribution from the same endpoint.[1][2]

Report an error or suggest a correction