Remember when we were all worried that AI would replace delivery drivers? Well, the future arrived faster than expected — just not the way anyone predicted. DoorDash didn't automate their workforce. They recruited them for something far more valuable: training the very AI systems that might one day make them obsolete.
Welcome to the era of "data labor." It's not delivery. It's not driving. It's something entirely new — and DoorDash just built the infrastructure to monetize it at scale.
Last week, DoorDash quietly launched "Tasks," a standalone app that pays their 8+ million delivery couriers (they call them "Dashers") to complete assignments that directly train AI and robotic systems. The tasks are bizarrely mundane: film yourself washing dishes while wearing a body camera. Record yourself speaking in other languages. Photograph restaurant dishes and hotel entrances. Walk through a grocery store and document shelf layouts.
DoorDash's official line? The data "helps AI and robotic systems understand the physical world." Which is corporate speak for: we're building a dataset that could power the next generation of automation, and we're paying workers minimum wage to create it.
Meet the Tasks App: Where Minimum Wage Meets Machine Learning
Let's break down what this actually looks like in practice. A Dasher opens the Tasks app and sees a list of available assignments. Not delivery orders — data collection tasks. Each comes with specific instructions, time estimates, and payment.
Example tasks DoorDash has listed:
Body camera recordings: Strap on a camera and perform everyday activities — washing dishes, folding laundry, organizing a closet. The footage captures how humans physically interact with objects from a first-person perspective.
Language samples: Record yourself speaking sentences in Spanish, Mandarin, or other languages. These become training data for voice recognition and translation systems.
Visual documentation: Photograph restaurant entrances, hotel lobbies, retail displays. The images train computer vision systems to recognize and navigate commercial spaces.
Environment mapping: Walk through spaces while recording video, creating the kind of spatial data that autonomous systems need to understand physical layouts.
The pay? Reports suggest anywhere from a few dollars per task to $15-20 for longer assignments. It's not much — certainly less than most Dashers could make delivering food during peak hours. But that's not the point. The point is that DoorDash has found a way to extract value from their workforce even when there aren't delivery orders to fulfill.
The Economic Loop From Hell
Here's where it gets philosophically uncomfortable. DoorDash has created what we might call "the training loop" — a circular economic relationship where workers are paid to accelerate the technology that could eliminate their current jobs.
Think about it: A Dasher delivers food for $5 plus tip. During slow periods, they open the Tasks app and earn $8 to film themselves performing household chores. That footage becomes training data for domestic robots. Those robots eventually become good enough to handle tasks that humans currently do. The Dasher's future employment prospects shrink — but hey, at least they made eight bucks.
It's not just DoorDash. This is becoming a pattern. Uber announced a similar program last year, offering drivers AI data-labeling gigs between rides. Amazon's Mechanical Turk has been the original "data labor" platform for years, paying pennies for tasks that train AI systems. What's new is the scale and the direct connection to a company's core automation strategy.
Why DoorDash Is Betting Big on Data
To understand why DoorDash is investing heavily in this, you need to understand their long-term automation play. This isn't just about training random AI models. It's about building the infrastructure for a fully automated delivery ecosystem.
DoorDash has already partnered with Waymo, Alphabet's self-driving car company. In some markets, Dashers are literally being paid to close the doors of self-driving delivery vehicles after retrieving orders. It's a strange intermediate step — humans assisting robots in the transition to full automation.
The Tasks app extends this strategy. While Dashers are filming themselves washing dishes or navigating retail spaces, they're creating the training data that could power:
Autonomous delivery robots: The kind that roll down sidewalks and navigate apartment building lobbies. They need to understand how to operate elevators, avoid obstacles, and interact with human spaces.
Warehouse automation: DoorDash has been expanding into grocery delivery, which means competing with Amazon in logistics. The visual data from Tasks could train systems for automated picking, packing, and sorting.
Computer vision for quality control: Identifying damaged goods, verifying order accuracy, monitoring food preparation — all tasks that currently require human oversight but could eventually be automated.
Voice and language systems: For customer service automation, multilingual support, and hands-free interaction with delivery systems.
The Dark Side: Exploitation or Opportunity?
Let's not mince words: There's something deeply uncomfortable about this arrangement. DoorDash has a history of controversial labor practices — classifying workers as independent contractors to avoid benefits, using algorithmic management to optimize worker behavior, paying rates that often work out to less than minimum wage after expenses.
The Tasks app adds a new dimension to these concerns. Workers aren't just being underpaid for delivery labor. They're now being underpaid for data labor — creating assets that have potentially enormous value to DoorDash and its partners.
The exploitation argument goes like this: Dashers are creating training datasets that could be worth millions or even billions of dollars in the long run. They're being paid pennies on the dollar for this work, with no equity, no royalties, no ownership of the data they create. It's extraction disguised as opportunity.
The counterargument: DoorDash is offering additional income opportunities during periods when workers would otherwise be idle. No one is forced to do Tasks. It's voluntary additional work. And the rates, while low, are competitive with other micro-task platforms.
The Uber Parallel: Following the Leader
DoorDash didn't invent this model. They're following a playbook that Uber started writing last year.
Uber's AI data program works similarly: Drivers opt in to data collection tasks during downtime. They might be asked to photograph street signs, verify business locations, or record interactions with passengers (anonymized, of course). The data trains Uber's mapping systems, their autonomous vehicle efforts, and their customer service AI.
The key difference is scale. Uber has roughly 5 million drivers worldwide. DoorDash has 8+ million Dashers. Combined, these two companies have access to a global workforce that dwarfs any traditional data labeling operation.
🔥 Our Hot Takes
The End of the Pure Gig Economy. We're witnessing the end of the gig economy as we knew it. It wasn't sustainable to pay people $5 per delivery forever. The new model is: Pay them for delivery when there's demand, pay them for data when there's not.
Data Labor Is the New Minimum Wage. As traditional gig work gets automated, "data labor" becomes the fallback employment. It's flexible, accessible, requires no special skills — and pays just enough to keep people participating without actually providing a living wage.
The Value Extraction Is Massive. Every video a Dasher records for $8 could be worth thousands of dollars in training value to DoorDash and their partners. The economics of this are wildly imbalanced.
Regulation Is Coming. At some point, regulators are going to notice that millions of workers are creating valuable intellectual property without any ownership stake or fair compensation.
The Irony Is Thick. Workers are literally training their replacements, and they're being paid poverty wages to do it. If that's not late-stage capitalism, I don't know what is.
The Bottom Line: A Preview of the Automated Future
The DoorDash Tasks app isn't just a quirky side hustle for delivery drivers. It's a preview of how the labor market evolves when AI starts eating jobs.
Here's the trajectory: First, AI assists workers. Then, workers train AI. Then, AI replaces workers. The Tasks app sits in that middle phase — the training period — where humans are still necessary but increasingly devalued.
For DoorDash, it's brilliant business strategy. They get to build valuable training datasets at rock-bottom prices, develop automation capabilities that will eventually reduce their labor costs, and position themselves as offering "additional earning opportunities" to their workforce.
For workers, it's more complicated. The Tasks app does offer additional income during slow periods. But it also accelerates the automation of the very jobs that currently pay their bills. Every video they record, every photo they take, every voice sample they provide brings the autonomous delivery future closer — a future where they might not be needed at all.
The $100 billion question isn't whether this model works. It clearly does. The question is whether we can build a society where the workers creating the data that powers automation get a fair share of the value they create.
Right now, the answer appears to be no. And that's a problem that extends far beyond DoorDash.
📚 Deeper Reading
Want to dive deeper into AI, tech, and productivity? Check these out:
- Deep Learning — The definitive textbook on deep learning by Goodfellow, Bengio, and Courville
- The Coming Wave — Mustafa Suleyman on the tsunami of technological change ahead
- Kindle Oasis — Premium e-reader with adjustable warm light and page turn buttons
- AI Superpowers — Kai-Fu Lee's insight on the AI race between China and the US
(As an Amazon Associate, we earn from qualifying purchases. These links help support AgentBear Corps.)