What is Agentic Commerce? The readiness checklist for businesses

You're chatting with your partner about hosting a dinner party. ChatGPT's ears perk up as you discuss the menu. The AI suggests recipes based on your dietary preferences, what it knows about the previous restaurants you've ventured to with Mark and Suzie & Greg and Jenny. It checks your pantry inventory through your smart home system, compares prices across Countdown, New World, and Farro, and then, without you visiting a single website, completes the purchase and schedules delivery for tomorrow morning.

No browser tabs, shopping carts, or checkout forms. Just a conversation that ends with groceries on their way.

This isn’t some far flung science fiction. It’s agentic commerce, and it’s already here.

Walmart partnered with OpenAI to enable exactly this scenario. PayPal integrated as the first payments wallet in ChatGPT. Shopify connected over a million merchants to conversational commerce. And behind the scenes, Google, Visa, and Mastercard each launched competing infrastructure protocols to enablement payments for this new era of shopping.

But what exactly is agentic commerce, and why is it a hot topic?

At its core, agentic commerce is shopping powered by autonomous AI agents that act on your behalf: researching products, comparing options, negotiating prices, and completing purchases, often without you explicitly commanding each step.

The term “agentic” comes from “agent”: an autonomous entity that can perceive its environment, make decisions, and take actions to achieve specific goals. Unlike traditional chatbots that only respond to your queries, agentic AI is:

Think of it as the difference between a search engine (you do all the work) and a personal shopper (they do the work for you). Except this personal shopper has instant access to every store globally, never sleeps, remembers everything you’ve ever purchased, and can negotiate on your behalf at scale.

1. Delegation of Authority: In traditional e-commerce, you control every step. In agentic commerce, you delegate authority to an AI agent. You might say:

The agent handles everything: research, comparison, vendor selection, price negotiation, payment, and confirmation. You define the goal and constraints; the agent figures out how to achieve it.

2. Multi-Agent/Multi-Vendor Orchestration: Traditional shopping happens in silos. Agentic commerce operates across all platforms simultaneously, querying multiple retailers, comparing delivery times, factoring in loyalty points or Airpoints, and executing the purchase through the optimal channel.

3. Contextual and Conversational: Agentic commerce is embedded in your daily life. You might say: You: “I’m training for a marathon”. Agent: “Based on your running history, you’ll need new shoes soon. Should I look at options? The ASICS store has your size on sale”. You: “Yes, but I need more cushioning this time”. Agent: “Got it. The Hoka Bondi 8 looks ideal. Would you like me to order them?”.

McKinsey expects agentic commerce to hit $1 trillion in the US alone by 2030, with a global range of $3–5 trillion. Citi Ventures projects 67% annual growth, pushing the category to $1.7 trillion by the same date. The behavioural signals are already here: retailers are reporting a 4,700% surge in agent-driven traffic, and customers arriving through agents show 2–6x higher purchase intent. Consumers aren’t fighting this shift either – 73% have used or would consider AI for product research, and 64% are open to using it for purchases, even if only 24% currently feel comfortable letting an AI complete the transaction end-to-end.

What we’re seeing is a structural reset in how online commerce works. If you’re not ready to accept agent-initiated transactions, your ability to compete online will erode fast over the coming years. Especially as customer traffic shifts away from websites and into AI channels; agents are becoming the new front door to discovery and shopping, and if you’re not integrated, you simply won’t be surfaced.

This isn’t a shift you scramble for in a few months. It requires rebuilding parts of your stack, rethinking trust, and aligning your data, payments, and product systems well before agents become the dominant channel.

Agentic commerce forces a total rethink of the payments stack. Today’s infrastructure is built on one assumption: a human is present, on a trusted interface, clicking “buy”. The moment an AI agent begins shopping, comparing, negotiating, and paying on your behalf, that assumption collapses.

But, autonomous transactions introduce risks the current system can’t handle: not because the tech is new, but because the trust model is completely different. When a machine is acting for a person, the system has three trust problems that we need to solve for.

These trust problems have been framed by Google as the 3As, and they explain exactly why the legacy stack breaks:

These are foundational gaps that make autonomous transactions unsafe. The existing rails were never designed to verify intent, nor to assign responsibility across humans and agents.

This is why the payments layer needs to be rebuilt.

To fix the trust model, the industry has begun designing new payment protocols that introduce cryptographic proof, verifiable identity, and clean execution paths for agents. Two major approaches have emerged, each solving a different layer of the stack.

1. Google’s Agent Payments Protocol (AP2)

payment rail that ensures any agent-led transaction has provable legitimacy.

At its core, AP2 uses Cryptographic Mandates. These are tamper-proof digital contracts that define exactly what the user has authorised.

These mandates generate Verifiable Digital Credentials that anchor the identity of both the user and the agent. They also produce immutable audit trails that make every step accountable.

AP2’s entire job is to solve the 3As:

2. OpenAI + Stripe’s Agentic Commerce Protocol (ACP)

ACP is the execution and checkout layer built into ChatGPT, focused on making agent shopping easy and widely adopted. Where AP2 focuses on solving for trust, ACP is focused on solving for a seamless shopping flow, and is largely merchant focused.

ACP’s role:

ACP’s philosophy is “Don’t rebuild your payment system, let agents plug into it". It solves distribution, convenience, and UX, but not the deeper trust and authorisation problems.

Think of ACP as the “Apple Pay moment” for agent shopping: frictionless, plug-and-play, instantly usable.

AP2 and ACP are ultimately competing industry initiatives, led by arguably the two biggest players in AI.

The strategic decision of whether a business should build on the ACP or AP2 currently has no definitive answer. The choice is highly dependent on a business's strategy, existing technology stack, and its primary goals, whether maximising conversion or ensuring strict governance.

While the initiatives are competitive, representing parallel efforts by industry giants to define the standard, they are often seen as operating at different layers: AP2 defines the cryptographic layer of permission and ACP defines the execution layer for instant transactions, suggesting eventual interoperability is likely, given many companies like Etsy and Shopify support both ecosystems.

Regardless of the choice, the common mandate for businesses across both standards is the same.

Here is the same tightened version with bolded recommendations so each action is clearly scannable:

1. Rebuild your data for machine-transactability (GEO). Readiness starts with restructuring your data so products and policies are fully machine-transactable and readability to autonomous AI agents. This requirement is known as Generative Engine Optimisation (GEO) and it is especially critical for Discovery Brands that compete on specifications, price, and availability. This means shifting from human-readable web pages to API-first offers, where product data, pricing logic, promo eligibility, loyalty rules, and warranty terms are expressed as structured logic an agent can parse and evaluate instantly. If an agent cannot fully understand or compare your offer, it will filter you out. That creates a new kind of abandonment: you disappear before the customer ever knows you exist.

2. Provide real-time operational data to avoid silent disqualification. You must maintain real-time status feeds for inventory, availability, shipping cutoffs, returns, and total landed cost. In the agent economy, abandonment happens when an agent cannot instantly validate these parameters and drops you from its shortlist. For Discovery Brands, failing to supply structured, real-time data becomes an existential threat to visibility and consideration.

3. Prepare your systems for agent-native payments (AP2). Autonomous payments break the assumption that a human is initiating the click, so you must adopt standards like AP2 to replace implied human action with verifiable, cryptographic Mandates. Merchants must be ready to process Cart Mandates for human-present checkouts and Intent Mandates for delegated, human-absent purchases such as price-drop reorders or stock-triggered replenishment.

4. Make your backend flexible, payment-agnostic, and mandate-aware. Your systems must store and validate Mandate evidence for accountability, while supporting a wide range of payment types including cards, bank transfers, digital wallets, and stablecoins. If high-conversion chat commerce is a priority, you must also ensure compatibility with execution-layer systems like ACP.

5. Upgrade your security and governance for autonomous agents. Agents introduce new risks: prompt injection, lateral movement, goal hijacking, and misaligned actions. Traditional payment security is no longer enough. You need Know Your Agent (KYA) protocols to verify an agent’s identity and intent, and all agent signatures and Mandates must be time-bound, context-bound, and instantly revokable to prevent replay or runaway actions.

The next five years will reshape commerce more than the last twenty. Start preparing now.