Choose your location

    Guide for QSR and retail

    Agentic commerce

    Learn how QSR and retail brands can stay visible as AI agents transform the way customers discover and buy. Explore what agentic commerce is, how it works, what is already live, and how to get ready for this game changer.

    Being readable by AI will soon matter more than advertising.

    Last updated: July 2026

    What agentic commerce actually is

    Agentic commerce is when an AI agent acts on your customer's behalf by making the purchase, not simply helping them decide what to buy. The customer tells the agent what they want: order my usual: a pepperoni pizza, regular fries, chipotle sauce, and a lemonade. Under $20, delivered on Friday before 8:15 p.m. The agent finds it, adds it to the cart, checks out, and pays, all following the preferences the customer sets. The agent does the shopping.

    For QSR and retail businesses, this means that, over time, AI agents, not customers, will increasingly decide which menus, products, and brands get chosen, or prioritized. Success depends on whether your business can be discovered, understood, and trusted by AI agents, not just by human shoppers.

    Example of agentic commerce:

    7:05 PM — Kitchen Smart Speaker or Phone

    User: "Hey, let’s do Thai tonight for dinner. Keep it under $75 total and delivered by 8:00."

    Agent: "Checking Lime Leaf Thai—their current prep time gets food here by 7:45 PM. Based on your last order, I have a draft cart: Pad Thai with chicken, Green Curry, Chicken Satay, and Basil Fried Rice. Want me to run with that?"

    User: "Tweak it. Sarah’s off meat this week, so switch her Pad Thai to tofu. And make sure the green curry is mild—it was way too spicy last time."

    Agent: "Got it.
    Swapped to Tofu Pad Thai
    Set Green Curry modifier to Spice Level 1 (Mild)
    With your saved $5 Lime Leaf reward applied, delivery, tax, and a 15% tip, the total comes to $64.10. Want me to send it?"

    User: "Wait, throw in an extra side of peanut sauce for the kids."

    Agent: "Added extra peanut sauce (+$1.50). Final total is $65.75. Charging your preferred Visa and placing the order now."

    And it could be so much more than dinner:

    • Grocery shopping

    • Online retail

    • Travel reservations

    • Pharmacy refills

    • B2B purchases

    How an agentic order works, step by step

    The customer tells the agent what they want

    In simple words, with their preferences: budget, delivery time, type of food, any dietary restrictions or allergies.

    The agent searches

    The agent scans the channels it can access: delivery apps, websites, and businesses whose menus, pricing, and hours of operation are readable by AI.

    The agent reads the live menu or catalog

    Items, prices, availability, and allergens. If your data is stale or unreadable, your business is invisible or will surface inaccurate info.

    The agent orders and pays

    The customer selects the option they want and the purchase is completed within the conversation.

    Your store fulfills it

    Your store fulfills the order. The order arrives just like any other online order. From there, your team delivers the experience.

    Agents that buy from you are agentic commerce. Agents that work for you are something else, and we never blur the two.

    Agents that buy from you
    Agentic commerce

    The customer's agent discovers, chooses, and pays. That agent brings the order in. That is what this guide covers.

    Acting for the customer · ChatGPT, Gemini, Alexa, and whatever comes next

    Agents that work for you
    Deliverect AI

    A digital workforce that optimizes menus, resolves support issues, and runs operations for the business. Those agents run the shop.

    Working for your business · live today

    Why it matters

    Agentic commerce benefits

    For businesses
    Be there when the agent decides

    Reach the growing number of customers who discover, compare, and order through AI assistants, not just apps and search engines. Capture revenue that might otherwise go to a competitor. The customer is going to place that order either way; you want the AI agent to recommend you. If your menu and pricing data are not structured and LLM readable, your business is virtually invisible to an agent, no matter how good the food is.


    Stay visible through this shift. The behavior is already here: 39% of US consumers have shopped with AI, and 85% of them say it improved the experience.

    For customers
    Less friction, more control

    Customers save time by letting the agent handle the whole buying process. The agent cuts decision fatigue with personalized recommendations, keeps purchases within the budget and preferences the customer sets, and removes the browsing, comparison, and manual checkout. Their agent will shop somewhere. The question is whether it can read you.

    The plumbing

    The protocol landscape, in plain words

    Three names you will keep hearing. These are layers, not competitors.

    Google and Shopify

    UCP

    Universal Commerce Protocol, an open standard for the full shopping journey: discovery, checkout, payment and post purchase. The common language that allows any AI agent to transact with any merchant, without custom integrations.

    OpenAI and Stripe

    ACP

    Agentic Commerce Protocol is an open standard that allows autonomous AI agents to securely authenticate and complete checkouts on a user's behalf.

    The connection layer

    MCP

    Model Context Protocol, the way AI agents connect to real systems and tools in the first place. The protocols above are compatible with it. Connection is one layer; the transaction is another.

    Now that you know the basics

    Where agentic commerce really stands today

    Credibility in a hyped category comes from saying plainly what is real. So here it is.

    Real

    The standards exist

    Open protocols for agentic transactions are live, built by the largest platforms in commerce, and the ecosystem behind them is serious.

    Started

    Food ordering has begun

    Assistants can already find restaurants, read live menus and place orders through the first integrations, and reorder meals via delivery apps. It’s real, but not widely used: early rollouts, limited markets, adoption unproven.

    Not Yet

    Agentic ordering at QSR scale

    An assistant reliably ordering across thousands of locations, franchise networks, peak hours and local menus: that does not exist yet. Anyone claiming otherwise is selling.

    What chain scale will actually demand, beyond the basics:

    Brand loyalty programs that follow the customer, with the same functionality as on the brand's mobile app

    Personalization and upsell logic owned by the brand, not the LLM platform

    Gift cards

    Brand payment rules, e.g. splitting one order across two cards, or taking cash only below a set order value and only on pickup

    Capacity management at peak hours

    Promotions, offers, and bundles beyond the trivial

    None of this layer is hypothetical: operators run these capabilities today. What is missing is the agent side, and the infrastructure side that connects agents to restaurants and stores. Until agents and protocols can handle them, the timeline stays long.

    The four signals that would mean it is accelerating: the first major chain seeing a material volume of its overall sales mix flow through agents. Assistants connecting directly to restaurants and stores by default, rather than routing through marketplaces. The protocols adding food specific capabilities like live availability and modifiers. And the first real order volumes being reported, not projected. When these start landing, the timeline compresses. We track them in the radar below.

    The numbers

    Agentic commerce statistics 2026

    It is clear, when you look at the data below, that customers have widely adopted AI to discover, compare, and evaluate products before completing the purchase themselves. That trend appears to be rapidly accelerating.  The part that is just starting is an agent completing the purchase on its own. They are not the same thing, and the former is the on-ramp to the latter

    +393%

    Growth of AI referred traffic to US retail sites, year over year, Q1 2026


    Adobe Analytics · 1 trillion visits tracked · Apr 2026

    42%

    How much higher AI referred visitors convert vs other channels, a complete reversal from a year earlier


    Adobe Analytics · Mar 2026

    $3T to $5T

    Global retail spend agentic commerce could orchestrate by 2030


    McKinsey · The Agentic Commerce Opportunity · Oct 2025

    ~50%

    Share of US consumers who research with AI but still resist letting an agent buy for them. Trust is the bottleneck, not the tech

    Bain and Company · Agentic AI in Retail · Nov 2025

    Will agents replace the delivery apps?

    Agentic commerce does not replace any of your channels. It adds one.

    The marketplace question

    Do agents kill the delivery apps?

    History says no. Every new channel was announced as the death of the one before it: the web, then branded apps, then the aggregators. Each time, digital ordering got bigger and the channels were additive instead of replacing each other. Agentic ordering is the next channel, not the last one standing. The brands that win will not be the ones that pick a side; they will be the ones set up to take the order wherever it comes from.

    The app and loyalty question

    What happens to your app and your loyalty program?

    Both stay central. Your app remains the home of the direct relationship: loyalty, personalization, your best margins. Your customer's agent is one more door into the same house, and your loyalty data is exactly what makes "order my usual" possible in the first place. The point is not app or agent. It is that every door feeds the same operation and the same customer relationship you own.

    The money question

    What does an agentic order cost you?

    Honestly, the economics aren’t clear yet. The earliest integrations launched with no marketplace commission, and the open protocols are designed to connect agents to businesses directly. But standards can carry their own costs over time, and the models will keep moving. The line to watch on every announcement is the take rate. When it settles, we will say so here, plainly.

    For QSR and retail operators

    What agent-ready will actually mean

    When agentic ordering reaches scale, the brands that win will be the ones whose data and operations an agent can work with, across every location. That preparation can start now.

    1. Menus and catalogs readable by LLMs, everywhere. Items, prices, availability and allergens, structured and current across hundreds or thousands of locations, not just the flagship store.

    2. One connection, all channels. An order should land in the kitchen the same way whether it came from an app, a kiosk, a website or an assistant.

    3. Operations that handle peak hours. An agent placed order is worthless if the store cannot make it, route it and reconcile it at Friday night volume. The unglamorous half decides who profits.

    4. Margin that holds across the franchise network. Whoever the agent is, each location should keep its customer, its numbers and its margin, and offers should fit the store, not blanket the brand.

    ~1/3

    of retail homepage content is invisible to AI models today

    ~66%

    of product page content is actually readable by AI models, and that is the best case

    Adobe, 2026. For an operator, the menu or catalog is the product page.

    Menus that differ by channel

    Item codes that do not match

    Items selling out at 8pm

    Prices localized per franchise

    Readability is only the entry ticket. This is the daily reality at chain scale, and an agent trusts your data exactly once. A good POS can cover these basics. The non trivial layer is the one listed in the state of the market above: loyalty, brand owned personalization and upsell, gift cards, payment rules, capacity, and real promotions.

    The hardest edge: allergens. When an agent orders for someone with an allergy, the accuracy of your allergen data stops being a marketing question and becomes a safety one. That is the real weight behind "readable by machines."


    What to ask your team this quarter. Five questions that tell you where you actually stand:

    1. Can a LLM read our menu or catalog today, with prices, availability and allergens, or only a human with a screen?

    2. Are we displaying real-time availability across all channels for the LLMS?

    3. Do our item codes match across every channel and location, or does each channel speak its own language?

    4. Could we accept an agentic order from anywhere tomorrow, without a six month integration project?

    5. If an agent placed an order right now, would we know it came from an agent, and could we reconcile it?

    Where Deliverect stands

    Ordering is the easy half

    The market conversation is about how agents will place orders. The harder question is everything after: making the order at Friday night volume, routing it, keeping menus true across a franchise network, protecting the margin per store. That operational half is where we have spent years. It includes the brand grade layer described above: loyalty, personalization and upsell logic, payment rules, capacity management, and promotions, running today across our network. It is the half that will decide who actually profits from agentic commerce.

    And the same work answers the buyer side question operators ask us most: how do we get discovered and trusted by agents? Keeping menus, availability, and pricing structured, accurate, and current across every location, through one connection, is exactly what makes a business readable to an agent. The data that runs your operation is the data agents read. That is where discoverability starts.

    Our position is simple: whoever wins the agent, the restaurant should win the order. We are building for that: connecting the agent side to the infrastructure operators already run on, with the first proofs of concept in motion. And we will tell you honestly what is real as it happens.

    1.5B+

    orders powered for the world's biggest QSR brands

    95,000+

    committed locations across 74 countries

    30M+

    API calls processed daily

    The Radar

    Relevant moves, as they happen

    A running drawer of what actually matters in agentic commerce for food and retail. These are the first pieces we described above: real launches, early stages, none of them chain scale yet. If it is here, it changes something.

    Square · Jul 2026

    ChatGPT and Claude ordering opens for US restaurants

    Assistants find restaurants, read live menus and place orders, with no marketplace commission.

    Google and Shopify · Jan 2026

    UCP launches as the open standard for agentic commerce

    One common language for agents, merchants and payments, endorsed across the ecosystem.

    Google · 2026

    Gemini reorders meals on DoorDash and Uber Eats

    Agentic task automation reaches food delivery on flagship devices.

    DoorDash · Jul 2026

    Real agent ordering opens to developers

    Early access lets AI agents place actual food orders through ordering infrastructure directly.

    Join us for exclusive tips

    At the end of each month we'll send you news, tips & tricks, and operational guides on how to make the most of your online sales channels.

    Questions people (and AI assistants) ask

    Meanwhile, agents already work for restaurants. Today.

    That is Deliverect AI: an army of virtual employees running menus, support and operations, live right now. Different thing, deliberately.