Why AI Agents Are Failing at DoorDash, But Hold Promise for Complex US Tasks

AI agents are demonstrably underperforming for simple tasks like ordering food on DoorDash, signaling a critical miscalibration of their current utility for US consumers. This minimal adoption suggests that while these intelligent systems may struggle with mundane, low-friction activities, their true value for American businesses and individuals lies in augmenting complex decision-making and sophisticated automation where human cognitive load is high.

DoorDash CEO Tony Xu recently revealed a stark reality for AI agent integration: despite their availability, these agents have generated only "several hundred orders" on the platform. This figure stands in stark contrast to the "just shy of 1 billion orders" DoorDash processed in the third quarter of 2026 alone, as shared during an earnings call. Xu candidly described this as a "very low use case,&quot acknowledging that when it comes to straightforward tasks like ordering a meal, users "tend to see that come directly to DoorDash." This data challenges the prevailing narrative that AI agents are destined to streamline every aspect of daily digital life, forcing a crucial reassessment of where and how they can truly deliver value.

The DoorDash Data: A Reality Check for Consumer AI Agents

The numbers from DoorDash are a sobering indicator for the immediate future of AI agents in consumer-facing applications, particularly for simple transactions. While large language models (LLMs) and autonomous agents have captured significant attention, their practical adoption for routine tasks highlights a disconnect between technological capability and genuine user need.

Tony Xu’s specific quote, "AI agents are generating several hundred orders [on DoorDash] across the time that they’ve been up and running," paints a clear picture. For a platform serving hundreds of millions of users, this usage is statistically negligible. It underscores a fundamental truth about human-computer interaction: for tasks that are already quick, intuitive, and low-friction, adding an intermediary layer—even an intelligent one—often introduces more complexity than it removes.

For US consumers, the process of opening an app like DoorDash, browsing options, and making a selection is already optimized for speed and convenience. The cognitive load required is minimal. Introducing an AI agent into this flow can feel cumbersome, requiring users to articulate preferences or confirm choices that they might otherwise make instinctively. This added layer can negate the perceived time-saving benefits, especially when users enjoy the act of browsing and discovering food options themselves.

Why AI Agents Struggle with Simple Transactions

The minimal uptake on DoorDash isn’t an isolated anomaly; it reflects several inherent challenges when applying AI agents to mundane, low-complexity tasks that US consumers frequently perform. The core issue often revolves around the perceived friction and lack of control.

• Increased Friction: For tasks like ordering food, most consumers prioritize speed and direct interaction. An AI agent, by nature, acts as an intermediary. Even a highly advanced agent might require clarifying questions or confirmation, adding steps to a process that’s already streamlined within native apps.
• Loss of Direct Control: Many users enjoy the tactile and visual experience of browsing menus, comparing items, and making spontaneous decisions. Delegating this to an AI agent can feel like losing control over a personal choice, particularly when taste and preference are highly subjective.
• Diminished Serendipity: AI agents are designed to optimize and execute. However, for simple purchases, users often appreciate the serendipitous discovery of new restaurants or dishes, something an agent focused on efficiency might overlook or deprioritize based on predefined criteria.
• Data Privacy Concerns: While users might trust DoorDash with their order history, integrating a third-party AI agent introduces another layer of data sharing, potentially raising concerns about how preferences and personal information are handled across different platforms.

The expectation that AI agents would seamlessly integrate into every aspect of our digital lives overlooks the human element of enjoyment and personal agency in simple daily tasks. "When it comes to a task that’s as simple as ordering food, we tend to see that come directly to DoorDash," Tony Xu noted, highlighting this innate user preference.

The True Frontier: "Long Consideration Purchases" and Complex Automation

While AI agents falter in the realm of instant gratification, their real promise, as Xu suggested, lies in "long consideration purchases"—scenarios where agents are "doing work on your behalf proactively in the background." This is where the US market can truly capitalize on AI agents’ capabilities: in complex, data-intensive, and high-stakes decision-making environments.

Consider the following areas where AI agents could become indispensable for US consumers and businesses:

Here are specific examples of complex tasks:

• Financial Planning and Investment: An AI agent could analyze investment portfolios across platforms like Fidelity and Vanguard, monitor market trends, compare intricate mortgage rates from multiple US banks, and suggest personalized retirement strategies, potentially saving thousands of dollars.
• Travel Itinerary Optimization: Instead of manually searching Expedia, Google Flights, and individual airline sites, an agent could orchestrate complex multi-leg trips, considering factors like flight prices, hotel amenities, car rental availability, visa requirements, and local transportation—all while optimizing for cost, time, or carbon footprint.
• Enterprise Procurement: For US businesses, AI agents could revolutionize supply chain management by autonomously identifying and vetting new suppliers, comparing contract terms, negotiating prices, and monitoring compliance across hundreds of vendors, far beyond human capacity.
• Healthcare Navigation: An agent could help patients navigate the labyrinthine US healthcare system, finding specialists, comparing insurance plan benefits, scheduling appointments, and even assisting with claims processing, easing a significant burden for many Americans.

In these scenarios, the AI agent doesn’t just execute a simple command; it performs extensive research, aggregates vast amounts of disparate data, applies sophisticated analytical models, and presents optimized solutions. This is where AI agents transition from a novelty to a genuine productivity enhancer, performing tasks that are too time-consuming, too complex, or too tedious for a human to manage efficiently.

Implications for US Businesses and Tech Development

The DoorDash data offers a crucial lesson for US businesses investing in artificial intelligence: calibrate AI agent development to solve *hard problems* where their unique capabilities truly shine, rather than forcing them into already optimized simple workflows. This means a strategic pivot from generalist "personal assistants" to specialized "expert agents" designed for high-value, high-complexity domains.

For startups and established tech giants alike, the focus should shift towards building agents that can: process unstructured data at scale, perform multi-step reasoning, integrate with diverse enterprise systems (like Salesforce or Oracle), and operate with a high degree of autonomy in defined, complex contexts. Venture capital funding should prioritize solutions that demonstrate clear ROI in areas like financial optimization, advanced logistics, scientific research, or specialized legal assistance.

Furthermore, as AI agents delve into more sensitive and impactful domains, regulatory bodies in the US, such as the Federal Trade Commission (FTC) and the Securities and Exchange Commission (SEC), will likely intensify scrutiny on issues like data privacy, algorithmic transparency, and accountability. Developers must build agents with robust ethical guidelines and audit trails, ensuring they can stand up to regulatory review and consumer trust demands.

Ultimately, the limited adoption of AI agents for simple consumer tasks shouldn’t be viewed as a failure of AI itself, but rather a clarifying moment. It underscores that the true transformative power of these technologies for the American economy and workforce will come from their ability to augment human intelligence in areas of significant complexity and strategic importance, freeing up human capital for creative and higher-level problem-solving.

Key Takeaways for AI Agents in the US Market

• AI agents show minimal adoption for simple consumer tasks like food delivery, as evidenced by DoorDash’s data.
• Their current value is low in scenarios with low cognitive load and established, efficient manual processes.
• The true potential of AI agents lies in "long consideration purchases" and complex automation, such as financial planning, travel optimization, and enterprise procurement.
• US businesses should focus AI agent development on specialized, high-complexity problems that significantly reduce human effort and improve outcomes.
• Regulatory oversight for data privacy and algorithmic transparency will become increasingly critical as agents handle more sensitive information and decisions.

Why aren’t US consumers using AI agents for food delivery?

US consumers are not widely adopting AI agents for food delivery because the process is already simple, fast, and highly optimized within native apps like DoorDash. Adding an AI agent often introduces unnecessary friction, reduces direct user control over personal choices, and removes the enjoyment of browsing, which many users prefer for mundane tasks.

What are "long consideration purchases" for AI agents?

"Long consideration purchases" are complex, multi-faceted decisions that require extensive research, data aggregation, comparison, and analysis over a prolonged period. Examples for US consumers include optimizing financial investments, planning intricate international travel itineraries, or navigating complex healthcare options, where an AI agent can significantly reduce cognitive load and improve outcomes.

How can US businesses best utilize AI agents?

US businesses can best utilize AI agents by deploying them for complex, data-intensive tasks that require significant human time and expertise. This includes automating enterprise procurement, optimizing logistics and supply chains, enhancing financial modeling and analysis, or personalizing customer service interactions for high-value clients, where agents can process vast information and identify optimal solutions efficiently.