For years, the modern restaurant owner has lived through a paradoxical nightmare. The digital dashboard shows record-breaking order volumes, yet the actual bank balance remains stubbornly flat. This is the delivery app trap, where the visibility provided by giants like DoorDash, Uber Eats, and Grubhub comes at a cost that often consumes the entire profit margin of a meal. The industry has accepted a silent tax on growth, where increasing sales often means increasing the amount of revenue handed over to a third-party intermediary. This week, the conversation in the merchant community shifted from how to survive these fees to how to bypass them entirely using the very technology that is currently redefining the white-collar workforce.
The Infrastructure of Agentic Commerce
Square has officially entered the era of agentic commerce by launching a suite of AI integrations that allow merchants to embed their storefronts directly into the world's most popular AI interfaces. By releasing a dedicated ChatGPT app and a Claude plugin, Square is transforming these LLMs from simple chatbots into active transactional agents. For the consumer, the experience is seamless. Instead of navigating a cluttered delivery app with endless scrolling and forced advertisements, a user can simply tell Claude or ChatGPT that they want a spicy vegan bowl from a local spot. The AI handles the discovery, the customization, and the order placement without the user ever leaving the chat interface.
This system is specifically designed for food and beverage sellers in the United States who have already activated their Square Online Ordering profiles. The brilliance of the implementation lies in its invisibility to the merchant. A restaurant owner does not need to write a single line of code or manage a complex API integration to participate. They continue to manage their menus, pricing, and inventory through the standard Square dashboard they already use. When an AI agent captures an order, that data is routed instantly to the Square POS (Point of Sale) and the KDS (Kitchen Display System). The order appears on the kitchen screen exactly like a walk-in or a phone order, ensuring that the adoption of cutting-edge AI does not disrupt the physical chaos of a dinner rush.
To solve the friction of payment, Square has implemented a dual-path checkout system. The most streamlined route allows users to complete the transaction immediately via Order by Cash App within the chat window. For those who prefer a traditional interface, the AI can redirect the user to the store's standard online ordering landing page. Crucially, this is not a blind redirect. The AI passes along every specific detail discussed in the chat—the extra toppings, the spice level, the dietary exclusions—so the digital shopping cart is already fully populated. The customer simply hits pay, turning a complex conversational interaction into a one-click conversion.
The Economic Pivot from Marketplace to Utility
To understand why this matters, one must look at the brutal math of the current delivery ecosystem. Traditional marketplaces charge a commission based on a percentage of the total order value, often ranging from 15% to as high as 30%. In this model, the more a restaurant sells, the more it pays the platform. Square is attempting to flip this script by removing the marketplace commission entirely. By leveraging ChatGPT and Claude as the discovery layer, Square eliminates the need for a middleman who charges for the privilege of being found. Instead, merchants pay only the standard online transaction processing fee, which sits at approximately 2.9% plus 30¢ per transaction.
This shift is complemented by a fundamental change in how delivery logistics are priced. While traditional apps often tie delivery costs to the order size or use complex dynamic pricing, Square has introduced a distance-based flat fee model. Rather than operating its own fleet of drivers, Square utilizes a white-label dispatch network. This infrastructure provides the logistics without the branding, allowing the restaurant to maintain its own identity. The resulting delivery fee is a fixed cost, typically between $7 and $10 depending on the distance. Because this fee is flat, it does not scale with the order volume. A $100 order costs the merchant the same in delivery fees as a $20 order, effectively protecting the profit margins on larger tickets and rewarding high-average-order-value establishments.
However, the most significant technical hurdle in AI commerce is the problem of hallucination. An AI that tells a customer a dish is available when the kitchen has run out of ingredients is a liability, not an asset. Square solves this through real-time catalog mapping. This technology creates a dynamic link between the store's live inventory and the AI's knowledge base. The AI agent does not rely on a static menu uploaded weeks ago; it queries the current state of the POS system. If a specific topping is marked as out-of-stock in the Square dashboard, the AI agent is instantly aware and will not offer it to the customer. This transforms the AI from a generic consultant into a highly trained digital employee who knows the exact state of the pantry in real-time.
By grounding the LLM in hard transactional data, Square has moved beyond the novelty of AI chat. They have built a system where the AI understands the constraints of the physical world. The result is an operational efficiency that mimics a human employee but operates at the scale of a global platform. The tension between the merchant's need for visibility and their need for profitability is resolved by replacing the expensive marketplace with a low-cost utility.
Commerce is moving away from the era of the centralized app store and toward a future of personalized AI agents. The power is shifting from the platforms that control the traffic to the systems that control the transaction.
