The Drone Can Fly. Who Loads the Burrito?

DoorDash has introduced a drone designed to deliver restaurant meals.

The company says its new aircraft can carry approximately 80 percent of the restaurant orders placed through its service. In preliminary testing, it reportedly transported meals from the restaurant to the customer’s door in less than five minutes.

That is impressive.

It may even be fast enough to deliver the french fries while they still possess some resemblance to french fries.

But DoorDash apparently discovered that building an aircraft capable of carrying lunch was not the entire problem.

Someone still has to load the burrito.

DoorDash Started on the Ground

When DoorDash recently announced DoorDash Air, much of the attention naturally went to the aircraft. It has six propellers, weighs less than 55 pounds and is designed to carry the smaller orders that make up most restaurant deliveries.

The company plans to begin testing the system in Northern California with restaurants including Chipotle and Popeyes.

But DoorDash says it did not begin by designing the drone. It began by studying the ground operation:

• How the meal would be packaged
• How it would move from the kitchen to the aircraft
• How the order would be loaded
• How restaurant employees would confirm that the correct meal was attached to the correct drone
• How the entire process would fit into an already busy restaurant

That may be the most important part of the announcement.

Flying the food may eventually be the easy part.

Getting it out of the kitchen could be considerably more complicated.

Restaurants Are Not Assembly Lines

A restaurant is not a controlled manufacturing environment.

Orders arrive at unpredictable times. Some customers want extra cheese. Others want no cheese, light sauce, extra sauce or something that does not appear to be available under the ordinary laws of food preparation.

The order must be cooked, assembled, packaged and matched with the correct customer. Drinks must remain upright. Soup must remain inside its container. Someone must notice that the dessert is still sitting on the counter.

At the same time, restaurant employees are serving customers in the dining room, processing drive-through orders, answering questions and handing meals to conventional delivery drivers.

Now one of them must also load the drone.

DoorDash says it is developing several loading systems that can fit into different restaurant operations. That is sensible because a Chipotle, a neighborhood pizza restaurant and a small Chinese takeout operation are unlikely to have the same space, staff or procedures.

The drone cannot simply land beside the cook and wait for someone to toss a burrito into it.

At least, I hope that is not the plan.

Automation Needs a Handoff

This is a recurring problem with automation.

The machine may perform its assigned task exceptionally well, but something must prepare the work before the machine begins and deal with the result when it finishes.

A warehouse sorting system can move thousands of packages, but people and other machines must place those packages into the system in the proper orientation. A self-checkout machine can process a purchase, but an employee still intervenes when it decides that a bag of tomatoes represents an unexpected item in the bagging area.

The DoorDash drone can carry lunch through the air. It cannot cook the meal, package it, determine whether everything is present or redesign the restaurant so employees can load it efficiently.

It also needs an acceptable place to leave the order.

A suburban house with an open driveway or front yard might be relatively straightforward. An apartment building, office complex, crowded sidewalk or neighborhood filled with trees, power lines and enthusiastic dogs creates a different problem.

The customer must identify an appropriate delivery area. The drone must reach it safely, lower or release the food, confirm delivery and depart without frightening the neighborhood—or placing someone’s chicken sandwich on the garage roof.

The flight connects two physical environments. Both must be prepared for the automation to work.

Five Minutes May Not Mean Five Minutes

DoorDash reports that its initial drone flights from the restaurant to the customer averaged less than five minutes.

That is certainly faster than driving through traffic.

But the customer is not particularly interested in flight time. The customer is interested in how long it takes for lunch to arrive.

The complete process includes:

• Accepting the order
• Preparing the food
• Packaging it for flight
• Moving it to the loading station
• Placing it in or on the aircraft
• Verifying the order
• Waiting for permission to depart
• Flying to the customer
• Completing the delivery

If the drone waits ten minutes for a restaurant employee to load it, the three-minute flight is less revolutionary.

The system therefore has to save time for the restaurant as well as for DoorDash and the customer. If it creates another complicated duty for already busy employees, restaurants may decide that a conventional driver standing at the counter is easier.

The limiting factor may not be aircraft speed. It may be the availability of the person carrying the bag out of the kitchen.

China’s Robots Have Discovered the Same Problem

The recent World Humanoid Robot Games in China produced impressive demonstrations of robots running, boxing, playing soccer and performing coordinated routines.

They also produced falls, collisions and other moments suggesting that human athletes do not need to begin looking for new work immediately.

The more meaningful events may have been the less spectacular ones. Robots were asked to perform activities resembling actual work: identify objects, manipulate tools, open containers, connect equipment and recover when something went wrong.

Those tasks do not produce the same headlines as a robot winning a race. They are much closer to determining whether anyone will pay the robot a salary.

Moving across a room is one capability. Picking up an unfamiliar object, using the correct amount of force and placing it precisely is another.

A robot that can run faster than I can is not especially unusual at this point.

Most people can run faster than I can.

A robot that can enter a cluttered hotel room, distinguish trash from someone’s belongings, replace the towels, make the bed and recognize a leaking faucet would be far more economically significant.

The robot games demonstrate how rapidly movement is improving. They also reveal how much more must be accomplished before general-purpose robots can perform useful work without creating additional work for the people around them.

The Human Job May Move

Drone delivery does not necessarily eliminate every delivery job. It changes where human work occurs.

Some driving may disappear, particularly for small orders traveling short distances. Other work may be created around:

• Preparing meals for drone transport
• Operating loading stations
• Maintaining and charging aircraft
• Monitoring flights
• Resolving failed deliveries
• Managing landing areas
• Complying with aviation and local regulations

Whether this produces more jobs, fewer jobs or simply different jobs remains uncertain.

What seems more predictable is that automation will first remove the portion of the process that is easiest to define. The irregular work will remain with people.

A drone can follow a programmed route. A restaurant employee must decide what to do when the order is ready, the loading station is occupied, the aircraft reports a problem and the customer has requested three additional containers of sauce.

Technology usually handles the routine part first.

People inherit the exceptions.

The Economics Must Include the People

A drone delivery system will be judged partly by whether it costs less than paying drivers.

That comparison can be misleading if the labor at the restaurant is treated as free.

If employees need additional time to package, verify and load drone orders, that time belongs in the calculation. So do aircraft maintenance, battery replacement, insurance, monitoring, ground equipment and the people needed when a delivery cannot be completed.

The correct comparison is not:

Drone versus driver.

It is:

The total cost of preparing, loading, transporting and delivering the order by drone versus the total cost of doing it by car.

The drone may eventually win that comparison. It uses less energy than moving a person and a 3,000-pound vehicle to deliver a two-pound meal. It avoids traffic and could complete several short deliveries in the time required for one trip by car.

But it wins only if the entire system works—not merely the part flying through the air.

The Last Fifty Feet

We tend to imagine automation replacing an entire activity at once.

More often, it replaces the cleanly defined portion and exposes all the untidy work surrounding it.

DoorDash may have created an aircraft capable of delivering food in five minutes. Now it must persuade restaurants that preparing and loading that aircraft will not disrupt everything else they are doing.

It must also deal with weather, noise, apartment buildings, trees, power lines, local regulations and customers who do not have an obvious place for a flying machine to leave lunch.

None of these problems means drone delivery will fail. DoorDash appears to understand that the ground system matters as much as the aircraft, which gives the effort a better chance than treating the drone as the entire solution.

The project also provides a useful reminder about automation.

The spectacular technology is frequently not the limiting factor.

The limiting factor is getting an ordinary person, working in an already busy restaurant, to place the correct burrito in the correct machine at the correct time.

And preferably remember the sauce.