Agentic Commerce in 2026: When AI Agents Become the Shopper
AI agents now shop, compare, and check out on people's behalf. Here is what changed in 2026, the money at stake, and how businesses stay on the shortlist.
Agentic commerce, shopping that an AI agent carries out on a person's behalf, moved from demo to real revenue in 2026. AI and agents already drove 20% of all retail sales and $262 billion in spending during the 2025 holiday season, according to Salesforce, and Bain projects agents will run $300 to $500 billion of US e-commerce, roughly 15% to 25% of the total, by 2030. The customer evaluating your store is increasingly an AI, or a human who trusts one to do the legwork.
This post covers what actually changed in 2026, how big the shift is on the best available numbers, the new payment protocols making it work worldwide, and what a business should do about it now.
What is agentic commerce, and what changed in 2026?
Agentic commerce is a purchase that an AI agent initiates, influences, or completes for a shopper. A person says "find me a cordless drill under 150 dollars with fast delivery," and the agent searches, compares, checks stock and price, and either hands back a short answer or buys outright. Bain draws the line deliberately: agentic commerce means agents that transact, not just AI-assisted search or discovery.
What made 2026 the turning point is that the big surfaces went live. Shopping assistants inside general assistants like ChatGPT, Gemini, Claude, Perplexity, and Copilot moved from answering questions to guiding whole purchases, and retailers deployed their own branded agents on-site. The traffic quality is the tell: shoppers referred to retailer sites from AI-powered search channels converted nine times more often than those arriving from social media, per Salesforce, and retailers that ran their own AI agents grew sales 59% faster than those that did not (6.2% versus 3.9% year over year).
The path is not a straight line. OpenAI launched in-chat Instant Checkout with Etsy and Shopify merchants in late 2025, then in early 2026 refocused the experience so that products still surface inside ChatGPT but buyers often complete the purchase on the merchant's own storefront. The surfaces and standards are still moving. The direction is not in question.
How big is agentic commerce, really?
Every major forecast lands in the same territory: agents will touch a large, growing slice of online spend by 2030, and adoption is already visible today. Because these are projections from different definitions, read the range, not any single number.
| Source | Forecast for 2030 | Adoption signal today |
|---|---|---|
| Bain & Company | $300 to $500 billion of US e-commerce, 15% to 25% | 30% to 45% of US consumers already use generative AI to research and compare products |
| Morgan Stanley | $190 to $385 billion of US e-commerce, 10% to 20% share | ~23% of Americans bought something via AI in the past month |
| McKinsey | Up to $1 trillion in orchestrated US B2C retail revenue, $3 to $5 trillion globally | About half of consumers now use AI somewhere when they search |
The consistent read across all three: within a few years, a meaningful minority of online purchases will be shaped or completed by an agent, and the behavior that feeds it (using AI to research and compare) is already mainstream. Morgan Stanley notes that spec-driven, repeat categories like groceries and consumer packaged goods lead the way, because those are the purchases shoppers most want to hand off.
The new plumbing: agentic commerce and payment protocols
The reason 2026 worked is infrastructure, not smarter chat. A stack of open protocols and card-network products now lets an agent prove who authorized a purchase, prove the agent is a real buyer and not a scraper, and settle the payment on existing rails. We covered the broader protocol shift in what is new in AI agents in 2026; here is the commerce-specific layer.
| Standard | Backer(s) | What it solves | Since |
|---|---|---|---|
| Agentic Commerce Protocol (ACP) | OpenAI and Stripe | Executing checkout inside an AI assistant | Sept 2025 |
| Agent Payments Protocol (AP2) | Google, donated to the FIDO Alliance | Cryptographically signed "mandates" that prove a human authorized the exact purchase | Sept 2025 |
| Trusted Agent Protocol (TAP) | Visa | Lets a merchant verify an inbound agent is a legitimate buyer, not a bot | Oct 2025 |
| Agent Pay | Mastercard | Tokenized card credentials scoped to a single agent, so disputes trace back to it | Apr 2025 |
| Machine Payments Protocol (MPP) | Stripe and Tempo | Machine-to-machine payments, including stablecoins, for the infrastructure layer | Mar 2026 |
| Agentic Mobile Protocol (AMP) | Ant International (Alipay) | Agentic payments across wallets and merchants, built for Asia and cross-border | 2026 |
Sources: OpenAI, Google, Visa, Mastercard, Stripe, and Reuters. These compose: an agent can pass Visa's identity check, then present a Google mandate as proof of consent, then settle on a card or a stablecoin.
Is agentic commerce only an American story?
No, and treating it as one misreads the field. The US card networks and assistant makers set much of the early standard, but Asia is moving just as fast. Ant International, the Alipay operator spun out of China's Ant Group, introduced its own Agentic Mobile Protocol and, in July 2026, raised $1.2 billion to scale it across a network it says reaches 150 million merchants and 2 billion accounts. Alibaba has folded AI shopping and sourcing agents into its marketplaces, and UnionPay is building its own agentic-payment framework. Mastercard's Agent Pay reached cardholders across Singapore, South Korea, and other Asia-Pacific markets early in 2026.
For any business selling across borders or in multiple languages, that global spread matters: the agent talking to your product data may be running on a US, Chinese, or European stack, and it will not wait for you to translate.
What does agentic commerce mean for your business?
Two shifts land on every business that sells through a website or a conversation, whether you run a Shopify store, a clinic, or a service business.
First, you have to be readable by AI. When an agent, or an AI-assisted human, is doing the comparing, your product data, prices, availability, and answers have to be accurate and machine-legible, or the agent recommends someone it can read instead. This is the commerce face of the same discipline behind getting cited by assistants; we go deep on it in how to get your business cited by AI search. Stale feeds and inconsistent prices do more than lose one sale: agents recheck facts at purchase time, and a wrong price signals unreliability.
Second, you still have to convert the humans. Most purchases are not fully autonomous yet, and higher-consideration buys (services, appointments, considered products) still run through a conversation: a DM, a web chat, a comment, a call. AI-referred shoppers arrive with higher intent, so the business that answers instantly, in the customer's language, and actually books or closes wins the moment. That is the conversational commerce side of the shift, and it is exactly what Entagl's Chat Agent is built for: an always-on AI that answers from your real catalog and knowledge across WhatsApp, Instagram, Facebook Messenger, Telegram, web chat, and API, in 30+ languages, and books real appointments into a real calendar, with a human able to take over at any point.
The through-line is the same one behind Entagl's four agents, one brain: being present where the buying happens, answering with accurate data, and optimizing to booked, paying customers rather than clicks. As buyers hand more of the journey to agents, the businesses that keep their data clean and their conversations answered stay on the shortlist. The rest become invisible, the way sites that ignored search once did.
What the data does and does not say
The forecasts are projections, not receipts, and they disagree at the edges (Bain's high case is 25% of US e-commerce; Morgan Stanley's is 20%). Adoption is real but uneven: Bain notes that most shoppers still are not comfortable letting AI complete an end-to-end transaction on its own, so trust, not technology, is the near-term ceiling. Spec-driven and repeat purchases will flip first; considered categories like apparel and travel follow as trust grows. And the standards themselves are still settling, as OpenAI's mid-cycle change to Instant Checkout showed. The safe planning assumption is not "agents buy everything next year." It is "a growing share of shopping starts inside AI, so be discoverable to it and ready to convert the people it sends you."
FAQ
What is agentic commerce?
Agentic commerce is shopping that an AI agent carries out on a person's behalf, from research and comparison through to checkout, with minimal human input. It differs from a normal chatbot, which answers one question and leaves you to act, in that the agent takes the steps itself. Bain scopes it to agents that actually transact, not just AI-assisted discovery.
Will AI shopping agents replace my website or store?
Not in the near term. Agents pull from your product data, prices, and availability, so your store still has to exist and stay accurate, it just may be read by a machine before a human ever sees it. The bigger risk is being unreadable to agents, not being replaced by them. Most higher-consideration purchases still complete through a human conversation or your own storefront.
How do I make my business ready for AI shopping agents?
Start with clean, complete, current data: accurate titles, attributes, prices, and stock, kept in sync so facts match at purchase time. Make sure your business is discoverable and quotable by AI, and make sure the conversations agents and AI-referred shoppers start (in your DMs, web chat, and calls) get answered instantly and can actually book or close. Being invisible or slow to agents is becoming as costly as being invisible on search once was.
Is agentic commerce only for large retailers?
No. The protocols and assistant surfaces are open to businesses of any size, and small operators that sell through conversations (clinics, salons, local services, creators, small e-commerce brands) benefit most from answering fast and accurately across channels. What matters is readable data and a responsive, always-on presence, not headcount.
Where this leaves you
Agentic commerce is not a 2030 forecast you can defer. It is already routing real revenue, the payment rails are live across the US and Asia, and the retailers that prepared their data outgrew everyone else last peak season. The unglamorous playbook wins: keep your product and business data accurate and machine-readable, and make sure every conversation an agent or a curious buyer starts gets answered and converted.
See how Entagl answers, qualifies, and books across every channel so you stay on the shortlist as more shopping moves inside the answer. Book a 30-minute demo.
Sources: Salesforce 2025 Holiday Shopping Data (Jan 2026); Bain & Company, 2030 Forecast: How Agentic AI Will Reshape US Retail (Dec 2025); Morgan Stanley, Here Come the Shopping Bots (Dec 2025); McKinsey, The Agentic Commerce Opportunity (2026); Modern Retail; Reuters (Jul 2026); and the OpenAI, Google, Visa, Mastercard, and Stripe protocol announcements linked above. Figures are third-party projections and reflect the sources' methodologies as of publication.