Agentic Commerce in 2026: What Happens When AI Agents Start Doing the Buying

For most of the last twenty years, the person browsing your website and the person clicking "buy" were the same person. That assumption is starting to break. In early 2026, a growing share of online purchases are being researched, compared, and completed by software acting on someone's behalf. The shopper types a request into ChatGPT or Gemini, and an agent goes off, reads the options, and checks out. The human never touches a product page.

This shift has a name now. Agentic commerce is the practice of letting AI agents discover, evaluate, and pay for goods and services with little or no step-by-step human input. It sounds like a niche experiment until you look at where the money is heading. Gartner expects AI agents to channel more than $15 trillion in B2B purchases alone by 2028. If that forecast lands anywhere close, the way businesses sell is about to change at the root.

What Is Agentic Commerce, Exactly?

Agentic commerce describes transactions where an autonomous AI agent handles the work a buyer used to do by hand. The agent reads context, weighs trade-offs, negotiates where it can, and completes the purchase through a checkout system it can talk to directly. The defining trait is that the buying decision and the buying action both move off the human's plate.

It helps to separate this from two things people confuse it with. A product recommendation engine suggests items but still waits for you to click. A support chatbot answers questions but cannot move money. Agentic AI commerce goes further than both. The agent has a goal ("reorder our office supplies under this budget" or "find a CRM that integrates with our accounting tool"), permission to act, and a way to settle payment. That combination is what makes it new.

Right now the consumer side is the loud part. AI shopping agents are live inside ChatGPT, Google Gemini, Microsoft Copilot, and Perplexity, and they are completing real purchases for real people. But the consumer headlines understate where the structural change is largest, which is business-to-business buying.

The Protocols Quietly Rewiring How Buying Works

None of this works without plumbing. An agent cannot buy from a store it has no standard way to read or pay. That gap is being filled by a new layer of open standards, and the most important one so far is the Agentic Commerce Protocol.

OpenAI built the Agentic Commerce Protocol together with Stripe and released it as an open specification. It defines a common set of APIs that let an AI agent securely reach a merchant's product catalog, pricing, and checkout. The first place it showed up at scale was Instant Checkout inside ChatGPT, which launched with U.S. Etsy sellers and then began onboarding Shopify merchants. When an order comes through, the merchant accepts or declines it, charges the card through their existing payment provider, and handles fulfillment exactly as before. OpenAI takes a 4% fee on each completed Instant Checkout purchase, and PayPal has been wired in as a payment option alongside cards.

The reason the open-standard approach matters is friction. Because the protocol is open-sourced, a merchant can implement compatible endpoints without locking themselves to a single AI company's stack. Stripe is not the only player either. Google used its NRF appearance in January 2026 to introduce a Universal Commerce Protocol aimed at letting agents interact with merchant catalogs through one shared standard. Visa, Mastercard, and American Express are all building their own agent-payment rails. The race is less about who has the smartest shopping bot and more about whose pipes everyone agrees to use.

This video from Stripe digs into the payment infrastructure underneath agent-driven buying and why the protocol design choices matter:

Why B2B Is the Bigger Story

Consumer agents buying socks make for good demos. The serious money is in procurement. Gartner's projection that AI agents will handle roughly 90% of B2B purchases by 2028, moving over $15 trillion, points at something most companies have not priced in yet. Reordering supplies, renewing software, reconciling invoices, and renegotiating contracts are exactly the repetitive, rules-bound tasks an agent does well.

Agentic commerce b2b looks different from a person filling a cart. The buyer's agent connects to vendor systems, pulls live pricing, checks it against a budget and an approval policy, and can haggle through machine-to-machine negotiation before committing. Gartner also warned that by 2028 AI agents could outnumber human sellers by ten to one, while fewer than 40% of sellers will say those agents actually improved their productivity. Read together, those two numbers describe a market flooding with automated buyers faster than sellers can figure out how to deal with them.

The companies that win here will be the ones whose data an agent can trust and parse. An agent cannot evaluate a vendor whose pricing lives in a PDF attached to an email, or whose product specs are scattered across three outdated pages. Verifiable, structured, machine-readable information becomes the price of being considered at all.

What Agentic Commerce Means for Sales Teams

If your buyer sends an agent, who are you actually selling to? This is the question reps have not had to ask before, and it reframes a lot of standard playbooks. Forrester predicts that during 2026, one in five sellers will be pushed to answer AI-powered buyer agents with dynamically generated counteroffers delivered through their own seller-controlled agents. Negotiation starts to happen bot to bot, at a speed no human could match.

The role of AI agents in sales cuts both ways. On the buy side, agents are screening vendors and pushing for better terms. On the sell side, teams are deploying agents to qualify inbound demand, respond to agent inquiries instantly, and hold pricing logic that can react in real time. A rep who relies on charm and a slow follow-up email is poorly positioned for a buyer that decided in four seconds.

Your Data Becomes Your Storefront

When a human visits your site, good design and persuasive copy carry weight. When an agent visits, it ignores the hero image and goes straight for the facts: price, availability, specs, terms, integrations, reviews it can verify. If those facts are inconsistent or hidden, the agent moves on to a competitor whose information is clean. Your structured data is now doing the selling that a landing page used to do.

The CRM Has to Speak Agent

This is where the back office stops being optional. If a buyer's agent asks for a quote, the request has to flow into a system that can produce accurate pricing, check stock or capacity, and log the interaction without a human retyping anything. A platform like Axelio, which keeps customer records, quotes, deals, and invoicing in one connected place, gives a business the structured foundation an agent-driven sale needs. Disconnected tools and stale spreadsheets are exactly what break down when the buyer on the other end is software that expects a clean, immediate answer.

How to Prepare Without Overreacting

It is easy to read forecasts like Gartner's and either panic or dismiss them. Neither helps. A few grounded moves matter more than any prediction:

• Clean up your data first. Make sure pricing, product details, and availability are accurate, current, and stored somewhere a machine can read. This pays off even if agents arrive slower than expected.
• Publish structured information. Use clear product schemas, consistent specs, and plain pricing pages instead of locking key facts inside PDFs or images.
• Watch your AI referral traffic. Visits coming from ChatGPT, Gemini, Perplexity, and Copilot behave differently from search traffic. If you are not measuring them, you cannot tell when agents start mediating your sales.
• Decide your protocol stance. Look at the Agentic Commerce Protocol and the payment rails the card networks are shipping, and choose deliberately rather than waiting to be forced.
• Give your team agent-aware playbooks. Reps need to know how to respond when the counterparty is software, and where a human still adds value the agent cannot.

The Risks Nobody Should Wave Away

This is not a clean story of progress. Handing purchasing to autonomous software raises real questions. Fraud and authorization get harder when the thing clicking "buy" is not a person. Brand loyalty erodes when an agent optimizes purely for price and specs and never sees your marketing. There is platform risk too, since a 4% take rate on agent-routed sales is fine until that channel becomes the channel. And there is the uncomfortable middle period Gartner flagged, where agents flood the market before most sellers have the tooling or the data quality to handle them well.

The sensible posture is neither hype nor denial. Agentic commerce is already moving real transactions, the protocols are being set now, and the businesses that get their data and systems in order early will be the ones agents can actually do business with. The rest risk becoming invisible to a buyer that never loads their website.

Frequently Asked Questions

What is agentic commerce in simple terms?

Agentic commerce is when an AI agent researches, compares, and completes a purchase on a buyer's behalf, instead of the person doing each step manually. The buyer sets a goal and a budget, and the agent handles discovery, evaluation, and checkout.

How is agentic commerce different from a chatbot or a recommendation engine?

A recommendation engine suggests items but waits for you to click, and a chatbot answers questions but cannot move money. An agentic AI commerce system can take a goal, decide, and actually pay, so the decision and the action both leave the human's hands.

What is the Agentic Commerce Protocol?

The Agentic Commerce Protocol is an open standard built by OpenAI and Stripe that gives AI agents a common way to read a merchant's catalog and pricing and complete checkout. Because it is open-sourced, merchants can support it without locking into one company's infrastructure.

Is agentic commerce actually being used yet, or is it still a concept?

It is live. AI shopping agents inside ChatGPT, Gemini, Copilot, and Perplexity are completing real purchases, and Instant Checkout in ChatGPT already works with U.S. Etsy sellers and is onboarding Shopify merchants.

Why is agentic commerce b2b considered a bigger deal than consumer shopping?

The volume is enormous. Gartner expects AI agents to handle around 90% of B2B purchases and channel more than $15 trillion by 2028, because procurement tasks like reordering, renewals, and invoice reconciliation are well suited to automation.

How does agentic commerce change the role of AI agents in sales?

Buyers increasingly send agents to screen vendors and negotiate terms, so sellers respond with their own agents that qualify demand and adjust offers in real time. Forrester expects one in five sellers to answer buyer agents with automated counteroffers during 2026.

What does my business need to do to be ready?

Start with data quality. Make sure pricing, product details, and availability are accurate and machine-readable, publish structured information instead of PDFs, and track traffic coming from AI assistants so you can see when agents begin mediating sales.

Will agents hurt customer loyalty?

They can. An agent optimizing only for price and specs may never see your branding or marketing, which weakens the emotional pull a brand usually relies on. Competing on verifiable value and clean information matters more in that environment.

How do merchants get paid in agentic commerce?

The merchant still processes payment through their existing provider and handles fulfillment, accepting or declining each order. In ChatGPT's Instant Checkout, OpenAI takes a 4% fee on completed purchases, and PayPal is available as a payment option alongside cards.

What role does a CRM play in agentic commerce?

A connected CRM is what lets an agent inquiry turn into an accurate quote and a logged deal without manual re-entry. Platforms that keep customer data, quotes, and invoicing in one place give agents the clean, immediate answers they expect.

Are there standards beyond OpenAI and Stripe?

Yes. Google introduced a Universal Commerce Protocol at NRF in January 2026, and Visa, Mastercard, and American Express are each building agent-payment rails. The competition is largely about whose standards merchants and agents agree to adopt.

What are the main risks of agentic commerce?

The big ones are fraud and authorization when a non-human is buying, loss of brand loyalty under price-only optimization, platform fees on agent-routed sales, and a transition period where automated buyers arrive faster than sellers can adapt their systems.

Sources

• Digital Commerce 360 — Gartner: AI agents will command $15 trillion in B2B purchases by 2028
• Gartner — AI agents will outnumber sellers by 10x by 2028
• OpenAI — Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol
• Stripe — Developing an open standard for agentic commerce
• Agentic Commerce Protocol — open specification (GitHub)
• commercetools — Agentic commerce stats and enterprise guide 2026
• Digital Commerce 360 — OpenAI expands agentic commerce push