Last week in this column, we wrote about how a viral fake Reddit post exposed some of the very real issues with the restaurant delivery model, particularly high costs for restaurants and consumers and low wages for delivery drivers.
These problems are difficult to solve because using human labor to transport food from a restaurant to a customer’s home is simply expensive. And yet apps like DoorDash and Uber Eats have amassed too large of an audience for restaurants to ignore.
That hasn’t stopped some would-be innovators from trying to come up with another way. There are a number of robotics startups aimed at automating food delivery. Wonder believes its vertically integrated model is the solution. And there has also been a concerted, but mostly ineffective, push to get customers to order delivery directly from restaurants’ own websites and apps rather than DoorDash and Uber.
None of them have really worked yet. DoorDash and Uber Eats continue to grow revenues quarter after quarter with no signs of slowing. Consumers may complain about the prices, but they can’t seem to quit delivery.
In Silicon Valley, a startup called the Delivery Collective is throwing its hat in the ring. It believes it’s found a way to offer customers a restaurant selection that rivals the big guys, but at a lower cost. And it’s leaning into AI chatbots like ChatGPT, rather than a mobile app, to capture demand.
Delivery Collective was founded in 2023 by Bala Subramaniam, who led product within Instacart’s fulfillment business and then Amazon’s package delivery team. After taking a break from the tech world from 2021 to 2023, he went looking to recapture the thrill of working at a fast-growing startup.
During the pandemic, Subramaniam’s family got into the habit of ordering food delivery, and he noticed that the prices just kept going up. This was because restaurants were marking up their prices on delivery apps to offset the 20% to 30% commissions they typically pay. On top of that, there were delivery fees and tips. And yet, like many consumers, Subramaniam kept using the apps because they were easy and convenient.
But he began thinking about ways to do delivery differently.
“What I thought we could take a look at is, if you were to create a blend of what works really well for the consumer of third-party delivery, and what works really well for the restaurant with direct ordering, and put it together in a reimagined product ecosystem, what would that look like?” Subramaniam said in an interview.
He knew from his time at Instacart that the cost of a 30- 45-minute delivery was fairly fixed at $6 to $7. But there was another part of the business model he thought was ripe for disruption: the cost of acquiring restaurants.
Delivery apps, he said, have built huge sales teams whose job it is to keep their marketplaces stocked with restaurants and maintain relationships with those operators. This allows the apps to offer the wide selection that keeps consumers coming back. But it also creates costs that are passed along to restaurants in the form of commissions, and to customers in the form of markups.
Subramaniam believed that if he could find a more efficient way to build a restaurant inventory, he could charge operators less for delivery service. There would still be the costs of actually delivering the food, but restaurants could choose to pass those to the consumer, who would ultimately still be paying less than they would on the other apps, because there would be no markups.
To build such a marketplace, Delivery Collective is hoping to team up with POS suppliers. Companies like Toast, Square, NCR and PAR work with tens of thousands of restaurants. If Delivery Collective could partner with them, it could add restaurants in bulk, providing a solid national selection at a lower cost.
What’s in it for POS companies? In a pitch sent to suppliers, Delivery Collective says by banding together to create a “decentralized” delivery marketplace, POS providers can fight back against third-party delivery apps, which eat into POS transactions and revenue. DoorDash is also challenging them more directly with reservations, dine-in rewards and a rumored POS system of its own.
Delivery Collective orders would get routed through a POS like any other order, with the POS getting its regular cut, and the restaurant keeping the customer data, which they do not get from traditional delivery apps. Delivery Collective itself would make money by charging restaurants $1 per transaction.
In other words, it wants to be a POS- and restaurant-friendly alternative to the major delivery apps. It has already received interest from suppliers in joining its “consortium,” Subramaniam said.
“We don’t want the POS companies to see us like a DoorDash or Uber Eats,” he said. “We are an insider.”
The question, of course, is how this small startup will be able to generate the kind of consumer demand that will be needed to make it viable. Its answer: AI chatbots.
There is a belief that chatbots like ChatGPT and Claude represent the next frontier for online shopping. Instead of using Google to search for pizza or pants, the thinking goes, consumers will simply type their query into ChatGPT, and it will go find the best options, like a personal assistant. They will be able to pay for their goods within these chabots as well.
Delivery Collective wants to stake its claim as one of the first restaurant ordering apps on ChatGPT. Its app, called Bites, will allow customers to browse restaurants, order and pay within the chatbot. (See a demo video here.)
The restaurant doesn’t have to do anything: It is automatically opted into the Bites marketplace via the app’s relationship with POS suppliers. Orders will appear like any other online order, with the additional $1 fee, of course. Bites is currently in the review process to join ChatGPT’s recently opened app store.
DoorDash, Uber and probably countless other online ordering companies are no doubt developing similar apps as they look to capitalize on chat-based ordering. But Bites believes it will be a preferred option because it will offer the best selection and lowest prices, and AI will be optimized to find the best deals for users. It says customers can expect to pay as much as $10-$12 less on Bites than on one of the major delivery apps; their savings will even be calculated at checkout.
Bites is still bite-sized. Currently available only in the Bay Area, it has about 600 restaurants listed on its mobile app. It acquired them the old-fashioned way—knocking on doors—not through relationships with POS companies. It recently hit $10 million in gross marketplace volume.
Bites also has a fleet of about 50 couriers, but still relies mainly on drivers from other delivery apps to fulfill orders. It may tackle the driver issue later, Subramaniam said, but for now it is focused on other pieces of the delivery puzzle.
It is a very complicated puzzle, and we are skeptical that anyone will be able to loosen the hold the major delivery apps have over consumers. The world of AI commerce, meanwhile, is still the Wild West. There is probably a fortune to be made there, but it’s still too early and murky to say how.
At the same time, we certainly don’t discourage anyone from trying to improve the delivery model, and Delivery Collective’s approach is one of the more creative and grounded we’ve seen.