Truly human-like robots are still a long way off.  But simpler A.I. systems that automate packaging and delivery could bring an end to the dominance of brick and mortar retail, sooner than you think.


A few months ago, Amazon CEO Jeff Bezos quietly invested in Heartland Robotics, a Cambridge, Massachusetts startup that wants to build industrial robots to replace “empower” workers all across America.  Heartland’s founders include Rodney Brooks, the robophile M.I.T. engineer who has written that when it comes to the rise of the machines, “resistance is futile.”

Bezos’s investment was small and didn’t involve Amazon’s funds.  But it made the news, and it led at least one tech blogger to wonder:  “Are we going to see giant metal arms packaging up boxes at Amazon warehouses?”

The answer is almost certainly yes.  And almost certainly that won’t be the end of the drive for automation at Amazon.  In today’s economy, the ability to use technology to replace people or make them unnecessary seems to be one of the key factors that separates the thriving from the suffering and dying.

In Amazon’s case, automation also is going to be the only way to remove a key remaining advantage of brick and mortar retail, namely the immediacy of the shopping experience it permits.  That immediacy helps explain why, despite all the talk about its demise, brick and mortar shoppers still account for about 96% of retail purchases.

The delay period between online buying and delivery can be divided into two main phases.  In the first, the order is processed and then the product is located, packaged and transferred to the shipper.  In the second, the shipper brings the packaged product to your door.

The first phase is the one over which Amazon generally has the most control, and where the traditional kind of industrial automation can probably speed things up considerably.  In a fully automated warehouse, where robots trundle around reading bar codes and grabbing items from shelves, the time from order-taking to having the package ready at the warehouse loading dock could probably be reduced to half an hour or less.

But then there is the second phase, in which the package is taken by truck to a nearby airport, loaded onto a plane, flown to the shipper’s local distribution center, scheduled for pickup by drivers, and then put onto a truck for final delivery.  All this takes at least 12 hours for overnight service, and more usually 24 to 48 hours.  In other words, at its fastest it is still much slower, and considerably more expensive, than the process of driving to, and buying from, a brick and mortar retailer.

What would an e-tailer such as Amazon have to do to close this gap?

Like Wal-Mart, But Filled With Robots

First, it would have to build and stock its product warehouses – now often called “fulfillment centers” – within a short driving distance of its buyers, thus eliminating the most time-consuming phase of the current shipping paradigm.  That may seem a bit odd, since it represents a shift backwards from the original e-tailing ideal – super-efficient and centralized – towards the space-wasting, decentralized brick and mortar model of a Wal-Mart.  But barring the development of some super-high-speed yet inexpensive transport technology, fulfilling orders locally is going to be the only way to make deliveries within a period comparable to a brick and mortar buying experience.

This also happens to be the trend now in place.  Amazon in recent years has been building new fulfillment centers around the world wherever it can justify them, many of them close enough to major population centers to enable short-haul ground transport.  Often, when I track an Amazon order I’ve made, I note that it seems to go by truck from its New Castle, Delaware or Sterling, Virginia fulfillment center, i.e., it never goes airborne.

Most Amazon orders still travel hundreds of miles and require air transport.  There isn’t enough e-commerce demand – yet – to justify a switch to purely local order fulfillment.  But Amazon does seem to have that goal in mind.  One of its big initiatives recently has been to try to boost the demand for its fulfillment centers and shipping systems by inviting other vendors to use them.  That’s in addition to the general push, by Amazon and other e-tailers, to expand the range of what can be sold online, from laundry detergent to motor oil to perishable groceries.

If a highly automated fulfillment center could bring a purchased item from its storage location on a shelf to the loading dock in half an hour, then obviously the next goal would be to get it to the customer’s door in a similarly short interval.

Here lies the largest hurdle, for in the present shipping paradigm, packages arrive at a Fed Ex or UPS distribution center during the afternoon or night, are sorted into individual truckloads, and then are sent out near dawn on routes that will take their drivers at least several hours to finish.  The system is set up to handle packages arriving by air, and its time scale is on the order of days.

By contrast, a locally-based “immediate-fulfillment” system would have to somehow whisk any package from the fulfillment center to its destination without further ado.  And it is hard to see how this could work, unless one could replace the big, expensive Fed Ex and UPS trucks, and their human, expensive drivers, with a much cheaper and nimbler mode of transport.

Enter the carbot, or botcar, or deliverybot, or whatever people end up calling it.  This would be a cheap-to-operate autonomous vehicle capable of delivering small loads over short distances, i.e., within an hour or so of setting out from the warehouse.  Right now this sort of thing may seem no more than a gleam in the eyes of futurists, but the technology is basically there:  Wheeled autonomous vehicles with robust navigation and anti-collision systems are already used in the field by the U.S. military, and DARPA’s recent Grand Challenge series showed that such technology can be adapted for existing cars and urban areas relatively easily.

Amazon and other e-tailers obviously can’t count on being able to deliver their orders by robots any time soon.  Most likely at least a decade will have to pass before people are comfortable sharing the road (or shoulders? or bike lanes?) with autonomous vehicles, whose designers also will have to contend with new issues of theft – of cargoes as well as of the bots themselves.  E-commerce will need to have a much greater share of retail, to justify the deployment of this immediate-delivery technology.   But, again, that trend is already in place.  And since the deliverybot would mark the endgame in the grand battle of e-tail vs. brick and mortar retail, you can bet that people like Jeff Bezos are going to push for its introduction.

One lesson I think we’ll end up learning here is that truly lifelike machines, such as those depicted in the film A.I., are not the most important source of robotic social disruption right now – because we won’t know how to build such things compactly and cheaply any time soon.  Relatively simple, non-humanoid systems such as the deliverybot, on the other hand, are already feasible.  And by forcing the extinction of brick and mortar retail, they could rock our society as powerfully as any technology ever has.

POSTSCRIPT:  NYT article (9 Oct 10) on Google’s carbot tech:

POST-POSTSCRIPT (25 Mar 2012): Amazon now replacing fulfillment center workers with robots: