Steve Burns sees a strange contradiction at the center of modern food delivery.

The customer experience feels highly automated. A person opens an app, chooses a restaurant, places an order, and watches the delivery move across a map.

Behind that interface, however, the final step often still depends on a person driving a conventional vehicle to carry a relatively small amount of food to someone’s house.

Burns, the entrepreneur behind NoTip, believes that mismatch explains much of the industry’s cost.

The software has changed. The physical delivery model has not changed nearly as much.

His bet is that the next meaningful improvement in food delivery will come from redesigning the last mile around the job itself.

The Expensive Part Is Hiding Behind the App

Food delivery can appear reasonably priced when the customer sees a delivery fee of several dollars.

Burns argues that this figure does not show the full economics.

Restaurants may pay substantial commissions to large delivery platforms, and they can respond by raising menu prices inside those apps. Customers may then pay additional service charges, delivery fees, and tips on top of a meal that already costs more than it would at the restaurant.

Burns uses a simple example. A meal priced around $30 at the restaurant can approach $60 after the various costs of delivery are included.

The problem, in his view, is less about any single fee than the infrastructure required to support the existing system.

A gig worker may use a large SUV to deliver a pizza. Another driver uses a compact car. Each vehicle has its own fuel consumption, maintenance requirements, insurance exposure, and operating characteristics.

There is very little standardization.

Burns saw the opposite model earlier in his career. He founded Workhorse, which developed electric delivery vehicles, and worked around fleets serving companies such as UPS and FedEx. Those operations taught him how valuable purpose-built equipment and fleet consistency can become at scale.

Food delivery, he believes, is ready for the same thinking.

A Three-Pound Pizza Does Not Need a 5,000-Pound Vehicle

NoTip is designing its system around a lightweight electric road vehicle carrying a smaller delivery robot.

The larger vehicle handles the trip from the restaurant to the customer’s curb. Once it arrives, a ramp deploys and the robot carries the order toward the front door.

Burns describes it as a dual-stage approach.

The goal is not to invent a completely new customer behavior. People already understand food being carried from a vehicle to a porch. NoTip is trying to reproduce that successful process without requiring a driver for every order.

The road vehicle is deliberately stripped of many things a passenger vehicle needs.

It does not have to move people, so it does not need the weight and equipment associated with carrying them. Burns says NoTip’s three-wheel vehicle is classified as a motorcycle and is designed around small-package delivery rather than passenger comfort.

That difference is central to the economics.

Burns estimates the vehicle currently uses roughly three cents of electricity per mile. On a six-mile round trip, electricity would amount to about 18 cents.

The broader idea is simple: a delivery system should become cheaper as it scales, rather than carrying the inefficiencies of millions of individually owned vehicles.

Physical AI Has to Work in the Messy World

Burns is particularly interested in what he calls physical AI.

Generating text or software output is one thing. Moving through a driveway, identifying a front door, avoiding obstacles, handling different properties, and completing a delivery is another.

That physical environment creates the harder engineering problem.

NoTip’s robot has to deal with real houses rather than controlled demonstrations. A walkway may contain a lawnmower. A home may have steps. The route to the door may be narrow or irregular.

Burns does not expect the system to solve every possible edge case immediately.

The current robot can extend an arm to handle several steps. If it encounters a situation it cannot safely navigate, the order may be left at the closest practical location.

That reflects a principle Burns has carried through startups for decades. A useful system does not have to solve every theoretical problem before it starts operating.

It has to work reliably enough to improve the economics of the job.

Lower Cost Is Only Part of the Argument

NoTip’s name comes from its proposed pricing model. Burns says the company plans to charge a flat $6 delivery fee without restaurant commissions or customer tipping.

He believes that could make food delivery accessible to people who currently use it sparingly because of the cost.

But his argument extends beyond price.

Human delivery involves road accidents, distracted driving, injuries outside vehicles, and other risks that become largely invisible when someone taps a button on an app.

Automation introduces its own safety requirements, but Burns believes a lightweight vehicle specifically engineered for small deliveries can ultimately reduce both cost and exposure.

His larger point is that food delivery has spent years optimizing the digital experience while leaving the physical process relatively untouched.

NoTip is built around the idea that the next improvement will not come from another ordering interface.

It will come from changing what actually moves the food.

Want more Plugged In? Read more or listen on Apple Podcasts.