In December 2013, online retailer Amazon.com (Seattle, Washington) famously demonstrated its then up-and-coming Prime Air drone-based delivery system on US newsmagazine television show 60 Minutes. During the demonstration, a Prime Air "octocopter" drone picked up a package from a conveyor belt at an Amazon warehouse, flew several miles to a home, and landed on the home's back patio, where it dropped off the package in view of its eager recipient. The drone then flew away—presumably to return to the warehouse to pick up another package and repeat the process with another waiting customer.

At the time, Prime Air seemed to herald an impending total overhaul of the business of package delivery—an overhaul in which drones and other advanced robotic vehicles replace human package carriers and deliver all manner of goods to recipients cheaply and perhaps within a matter of minutes. Since that time, robotic package delivery has seen no shortage of industry activity: Start-ups, large logistics companies, and retailers alike have expended vast sums in the development of all manner of automated vehicles for carrying packages through the air, on sidewalks, inside buildings, and on public roadways.

But thus far, notable real-world examples of delivery by robot have been few and far between. And the few companies that have used robots to deliver actual packages to paying customers on an ongoing basis have built enterprises that have relatively little in common with Amazon's vision of rapid single-package, direct-to-end-user delivery. For example, logistics company JD.com's (Beijing, China) main drone-delivery enterprise uses large-capacity drones to fly many packages at once to dedicated hubs in rural locations, where JD personnel then unload the packages and perform actual last-mile delivery to the end user. The drone in use by JD performs the kind of long-haul delivery that has been the target of little automation effort thus far, but this situation is changing rapidly: Companies such as Daimler (Stuttgart, Germany), Tesla (Palo Alto, California), and Volvo Group (Gothenburg, Sweden) have demonstrated automated long-haul trucks that humans can supervise.

Meanwhile, many of the same companies that have been working to replace human package-delivery personnel with robots have also been using artificial-intelligence and related technologies to make human deliverers much more cost-effective, responsive, and versatile. These efforts challenge automated systems' supposed cost and performance advantages. Amazon still has not deployed Prime Air, but the company has nevertheless overhauled the process of package delivery through its Amazon Flex service, which enables individuals to sign up to deliver packages for Amazon directly. Under the guidance of the Flex smartphone app, Flex users drive to a local Amazon warehouse to pick up packages and then deliver the packages to customers' doorsteps directly. Flex has, in large part, helped Amazon gain the ability to offer services such as same-day delivery—the same sorts of services that its Prime Air drones would have provided. And Flex is just one of a great many similar services that match people who wish to deliver items from one point to another with companies that require delivery services. Many millions of people worldwide now deliver fresh food from local restaurants to hungry restaurant patrons' homes and offices and even to the restaurant patrons themselves (on the basis of the patrons' real-time locations). The abundance of such services has even led to the phenomenon of virtual restaurants, which have no customer-accessible locations and instead exist solely to send food directly to patrons through crowdsourced or dedicated delivery services.

Whether and to what extent robots transform the business of delivery depends on a great many forces and factors that have uncertain outcomes. A recent McKinsey & Company (New York, New York) study proposed that automated systems could come to deliver up to 80% of all packages, with bike couriers and conventional delivery methods making up the balance. But the study assumed that the primary automated last-mile delivery system would be a kind of automated mobile package locker that would drive up to the curb in front of a package recipient's building and then wait for the recipient to come outside and retrieve the package from the vehicle in response to a notification on his or her mobile device. Such a scheme would work well with the limitations of package-delivery robots, which currently can be fairly good at navigating the last mile by themselves but have incredible difficulty in navigating the last few meters from the street in front of a building to a recipient's doorstep; however, end users would have to change totally their behavior relating to package delivery, and all indications suggest that end users will not want to do so. Developments such as the proliferation of secure package lockers in apartment-building lobbies and the availability of Amazon Key and other products that permit a delivery person to drop off a package inside a recipient's home securely strongly indicate that end users do not want to have to meet their packages when they arrive. Instead, they want those packages to be waiting for them when they are ready. Only in well-defined cases, such as fresh-food delivery, does people's going out of their way to accept a delivery make sense. And services such as Cambridge Consultants' (Cambridge, England) DelivAir, which uses drones to deliver packages to wherever a recipient is in real time, promise to reduce the extent to which recipients must suffer inconvenience. Of course, DelivAir still has to compete with an army of ultra-low-cost humans who are willing and able to perform the same task, perhaps for the same price.

Using robots to automate the last mile of delivery faces many other challenges, including regulatory backlash—particularly from city governments that do not like the idea of drones' interfering with citizens' enjoyment of the city's sidewalks, roadways, and airspace. Because humans can use the same information technologies that robots can use, humans may continue to be the strongest challenge that robot-based delivery automation faces. New automation technologies promise to transform the way companies deliver goods to people, but humans may end up doing most of the work.