DoorDash turns Gig Workers into AI data Engines with New “Tasks” app

DoorDash is quietly redrawing the boundaries of the gig economy, launching a standalone “Tasks” app that converts its delivery workforce into a scalable pipeline for artificial intelligence training data—an asset increasingly viewed as more strategic than compute power itself.

At its core, the initiative is viewed as part of a broader structural shift in how AI systems are developed. While much of the first wave of generative AI relied on scraping vast amounts of internet text, the next phase—particularly robotics, autonomous systems, and “agentic” AI—requires grounded, real-world data. DoorDash’s network of millions of couriers offers precisely that: human-labeled, context-rich inputs generated in uncontrolled, everyday environments.

The Tasks app operationalizes this advantage. Couriers are paid to complete structured assignments—filming routine activities, capturing physical environments, or recording speech—which are then used to train models that need to interpret the physical world with high accuracy. The company says the data will support both its internal systems and those of external partners across industries, positioning DoorDash as a data infrastructure provider rather than just a logistics platform.

This evolution mirrors moves by rivals such as Uber, which has begun testing similar micro-task programs. The convergence points to a broader recalibration in the gig economy: platforms are no longer just intermediaries for labor and demand, but are becoming critical suppliers in the AI value chain.

What distinguishes DoorDash’s approach is its ability to integrate data collection directly into existing workflows. Tasks are embedded within the Dasher app, alongside delivery jobs, allowing the company to gather hyper-local, real-time data at minimal additional cost. This creates a feedback loop—data collected from the field can immediately improve route optimization, mapping accuracy, and customer experience, while also feeding longer-term AI development.

Analysts say this dual-use model could materially improve margins over time. Delivery remains a low-margin business, heavily exposed to fuel costs, labor incentives, and competition. Data, by contrast, scales with far higher profitability. If DoorDash can successfully package and sell AI training datasets—or embed them into higher-value enterprise services—it could open a new revenue stream less sensitive to the cyclical pressures of consumer spending.

It comes at a time when demand for high-quality training data is surging as companies like OpenAI and Google push toward more autonomous systems capable of acting, not just responding. These systems require “ground truth” data—accurate representations of real-world conditions—to function reliably. Synthetic data can fill gaps, but it often lacks the unpredictability and nuance of human environments.

DoorDash’s network effectively becomes a distributed sensor layer, capturing edge cases that are critical for AI performance. For example, variations in lighting, object placement, human behavior, or language accents—factors that are difficult to simulate—can be systematically recorded and fed into training pipelines.

There is also a geopolitical and competitive dimension. As governments tighten restrictions on cross-border data flows and companies guard proprietary datasets, access to unique, internally generated data is becoming a key differentiator. DoorDash’s model allows it to build such a dataset organically, without relying on third-party sources that may be restricted or commoditized.

However, the strategy introduces new tensions around labor and data ownership. Couriers are effectively producing high-value digital assets, yet are compensated on a per-task basis with no ongoing claim to the downstream value created. As AI models trained on this data generate revenue, questions around fair compensation, data rights, and transparency are likely to intensify—echoing earlier debates over how social media platforms monetized user-generated content.

There are also privacy and regulatory considerations. Tasks involving video, audio, or location data raise potential concerns about consent, data storage, and usage, particularly as the app expands into new markets with stricter data protection regimes.

Still, for DoorDash, the upside is clear. By leveraging an existing workforce to solve one of AI’s most expensive bottlenecks—data acquisition—the company is effectively lowering the barrier to entry for itself and its partners in the AI ecosystem.

The rollout remains limited to select U.S. markets, but the model is inherently global. With operations spanning multiple countries, DoorDash could eventually replicate the system internationally, creating one of the largest human-in-the-loop data networks in the world.