What Is Agentic Commerce? The Complete Guide for Merchants (2026)
Nearly half of all consumers already use artificial intelligence somewhere in their buying journey. Shopping-related AI queries grew 4,700% between 2024 and 2025. And McKinsey projects that AI agents could redirect $3 to $5 trillion in global retail spending by the end of the decade.
What Is Agentic Commerce? The Complete Guide for Merchants (2026)
Last updated: March 2026
Nearly half of all consumers already use artificial intelligence somewhere in their buying journey. Shopping-related AI queries grew 4,700% between 2024 and 2025. And McKinsey projects that AI agents could redirect $3 to $5 trillion in global retail spending by the end of the decade.
The question is no longer whether agentic commerce is coming. It is whether your business is ready for it.
This guide is the comprehensive resource for executives, merchants, and decision-makers who need to understand what agentic commerce is, how it works, who is building it, and what it means for the future of retail.
What Is Agentic Commerce?
Agentic commerce is a model in which artificial intelligence agents research, negotiate, and complete purchases on behalf of consumers or businesses.
Unlike traditional e-commerce, where a human browses a website, adds items to a cart, and checks out manually, agentic commerce delegates part or all of that process to an AI agent. The consumer expresses an intent – “find me running shoes under $150 that work for marathon training” – and the agent handles the rest: searching across merchants, comparing options, evaluating reviews, assembling a cart, and, when authorized, completing the transaction.
McKinsey describes it as “a seismic shift” that transforms shopping from a series of discrete steps into a “continuous, intent-driven flow.” As their October 2025 report states: “This is not just an evolution of ecommerce. It’s a rethinking of shopping itself.”
The shift is fundamental. In traditional e-commerce, the customer visits the store. In agentic commerce, the agent visits the store on the customer’s behalf. The buyer may never see a product page, never scroll through search results, and never manually enter payment details. The AI does the legwork.
This does not mean humans are removed from the process. Most agentic commerce today operates in a supervised mode where the agent recommends and the human approves. Fully autonomous purchasing – where an agent buys without any human confirmation – remains rare and limited to low-stakes, routine purchases. But the trajectory is clear: agents are moving from assistants to decision-makers.
How Agentic Commerce Works: The User Journey
A typical agentic commerce interaction follows four stages:
1. Intent Expression. The consumer tells an AI agent what they need, using natural language. This could happen in ChatGPT, Google Gemini, a WhatsApp conversation, a voice assistant, or any AI-powered interface. Example: “I need a birthday gift for my sister. She likes skincare, and my budget is $75.”
2. Agent Research and Evaluation. The agent searches across multiple merchants, evaluates product catalogs, reads reviews, compares prices, checks availability, and applies the consumer’s preferences and constraints. Unlike a search engine that returns a list of links, the agent synthesizes information and narrows the options.
3. Recommendation and Assembly. The agent presents a curated shortlist or a ready-to-purchase cart. It may explain trade-offs (“Option A is higher rated but $20 over budget; Option B fits your budget and ships faster”). The consumer reviews the recommendation.
4. Authorization and Checkout. The consumer approves the purchase. The agent handles checkout, applying payment credentials, shipping preferences, and any loyalty rewards. In more advanced implementations, the agent can execute the purchase autonomously based on predefined rules – for example, “buy it if it’s under $80 and arrives by Friday.”
This four-stage flow compresses what traditionally required dozens of browser tabs, multiple store visits, and hours of comparison into a single conversational interaction that takes minutes.
The Protocol Stack: How Agents Talk to Merchants
For agentic commerce to work at scale, AI agents need standardized ways to discover products, communicate with merchants, and process payments. A new protocol stack is emerging to make this possible.
| Protocol | Full Name | Created By | Purpose |
|---|---|---|---|
| UCP | Universal Commerce Protocol | Google + Shopify | Open standard for product discovery, checkout, and post-purchase flows between agents and merchants |
| ACP | Agentic Commerce Protocol | OpenAI + Stripe | Platform-mediated commerce enabling purchases inside ChatGPT |
| AP2 | Agent Payments Protocol | Secure payment authorization for agents acting on behalf of users | |
| MCP | Model Context Protocol | Anthropic | Standardizes how AI agents connect to external tools, APIs, and data sources |
| A2A | Agent-to-Agent Protocol | Enables communication and task delegation between multiple AI agents |
How they fit together. MCP defines how a single agent uses its tools. A2A defines how multiple agents collaborate. UCP defines how agents discover and transact with merchants. ACP provides a platform-specific commerce layer within ChatGPT. AP2 handles the payment authorization layer, ensuring agents can pay securely on behalf of users.
UCP and ACP represent two competing but complementary philosophies. UCP is decentralized: merchants publish their capabilities at a standard endpoint, and any agent can discover and interact with them. ACP is platform-mediated: merchants submit data to OpenAI, Stripe handles payments, and ChatGPT surfaces products to users. Both are open source. Both are live in production.
The important takeaway for merchants: these protocols are converging around a common goal – making your products discoverable and purchasable by AI agents, regardless of which platform the consumer uses.
Who Is Building Agentic Commerce
Agentic commerce is not a niche experiment. The world’s largest technology companies, payment networks, and retail platforms are building the infrastructure.
AI Platforms:
- OpenAI launched Operator in January 2025 and Instant Checkout inside ChatGPT in February 2026, enabling users to purchase products directly within conversations.
- Google released AI Shopping Mode with access to 50 billion product listings, co-created UCP with Shopify, and developed both A2A and AP2.
- Perplexity launched “Buy with Pro,” an agentic shopping tool, in late 2024.
- Amazon deployed Rufus, its AI shopping assistant, reaching 250 million users with 60% higher purchase conversion compared to non-AI sessions.
Commerce Platforms:
- Shopify co-created UCP with Google and enrolled over 1 million merchants into OpenAI’s commerce integration. Its Agentic Storefronts feature lets merchants publish product data once and have it surface across all major AI agents.
Payment Networks:
- Visa launched its Intelligent Commerce AI platform in partnership with eight companies including Anthropic, IBM, Microsoft, OpenAI, and Stripe.
- Mastercard and PayPal are developing their own agentic shopping and payment services.
- Stripe co-created ACP with OpenAI and serves as the payment processor for ChatGPT Instant Checkout.
Advisory and Research:
- McKinsey, Bain, BCG, Morgan Stanley, Gartner, and Forrester have all published major research framing agentic commerce as one of the defining shifts of the decade.
This is not a speculative technology waiting for adoption. It is an active buildout with production systems, real transaction volume, and major institutional investment – $9.7 billion in agentic AI startups since 2023, with infrastructure spending commitments exceeding $500 billion (Project Stargate). Agent orchestration platform headcount grew 87% year-over-year (CB Insights), and the surge in agentic framework usage between 2023 and 2025 reached 920% (Market.us).
The competitive dynamics are also worth noting. During Cyber Week 2025, AI agent-influenced sales reached $67 billion globally – 20% of all purchases (Salesforce). Retailers with branded AI agents saw sales grow 32% faster than those without. Amazon’s Rufus AI assistant is already pacing at over $10 billion in incremental annualized sales. These are not projections. They are current performance metrics from live systems.
Market Size and Projections
Every major research firm has issued projections for agentic commerce, and while the ranges vary, the direction is unanimous: this market is large and growing fast.
| Projection | Value | Source |
|---|---|---|
| Global agentic commerce by 2030 | $3 – $5 trillion | McKinsey (Oct 2025) |
| US B2C retail orchestrated by agents by 2030 | Up to $1 trillion | McKinsey (Oct 2025) |
| US e-commerce agentic spending by 2030 | $190B – $385B (10–20% share) | Morgan Stanley (Dec 2025) |
| US agentic commerce market by 2030 | $300B – $500B (15–25% of online retail) | Bain & Company |
| B2B spending through AI agent exchanges by 2028 | $15 trillion | Gartner (Nov 2025) |
| AI-driven retail spending in 2026 | $20.9B (nearly 4x 2025) | eMarketer (Dec 2025) |
| AI in retail/eCommerce market by 2030 | $175B at 30.2% CAGR | Wipro |
| Global AI agents market by 2034 | $236B at 45.8% CAGR | Precedence Research |
The B2B opportunity is particularly striking. Gartner predicts that by 2028, 90% of B2B buying interactions will be AI-agent intermediated, representing over $15 trillion in purchasing volume.
Consumer Adoption: Where We Stand Today
Consumer behavior data reveals a market in rapid transition – AI is already deeply embedded in the shopping journey, but full purchasing autonomy remains a frontier.
Usage is widespread and accelerating
According to an IBM-NRF study of 18,000 consumers across 23 countries (January 2026), 45% of consumers already use AI in their buying journey. Usage spans research (41%), review interpretation (33%), and deal hunting (31%). A BCG survey found that GenAI usage for shopping grew 35% in just nine months during 2025.
The platforms reflect this. ChatGPT handles an estimated 53 million shopping queries per day. AI traffic to US retail sites grew 805% year-over-year on Black Friday 2025 (Adobe). Amazon Rufus usage grew 127% in five months.
But trust in autonomous purchasing is still low
While consumers readily use AI for research and recommendations, they are far more cautious about letting agents handle transactions. A ChannelEngine study of 4,500 shoppers across five countries found that only 17% feel comfortable completing a purchase entirely through AI. Just 14% trust AI to place orders on their behalf (Clutch, January 2026). And only 4% would trust AI to complete a purchase without any final human review.
The gap is clear: 58% have used AI to research products, but only 13% have completed a purchase after being referred by an AI assistant. Payment security is the number-one concern, cited by roughly one-third of respondents across multiple surveys. Privacy compounds the issue – while 52% of consumers say they are comfortable sharing data with AI agents, 83% simultaneously express concerns about how that data is used (IBM-NRF). This “privacy paradox” – willingness and worry coexisting – is one of the defining tensions of the current moment.
What builds trust? Accuracy ranks first by a wide margin. Seventy-nine percent of consumers say accuracy is the most important quality in an AI shopping assistant, far outpacing speed (36%) and transparency (35%). Trust accounts for up to 60% of the explained variance in AI-assisted purchase intention (RealityMine/BCG). What destroys trust? Irrelevant recommendations are the fastest killer – 69% of early users immediately abandoned an AI assistant after receiving irrelevant product suggestions.
Generational divides are real but narrowing
Gen Z and Millennials adopt AI shopping tools at roughly twice the rate of Gen X and eleven times the rate of Boomers. Thirty-five percent of Gen Z and 32% of Millennials used ChatGPT for product search in the past month, compared to 23% of Gen X (Forrester, December 2025). Seven out of ten Gen Z shoppers now use AI to buy online (IESE).
However, the fastest growth is among older demographics – Boomer AI adoption surged 92% and Gen X adoption grew 82% year-over-year (IBM-NRF). This pattern suggests that agentic commerce is following a classic technology adoption curve, with younger cohorts leading but mass-market adoption accelerating rapidly across all age groups.
Geographic variation is also significant. GenAI usage for shopping in Brazil reaches 63% and in India 62%, compared to 42% in the United States and 48% in Japan (BCG). Emerging markets, many of which already have strong conversational commerce cultures through platforms like WhatsApp, may adopt agentic commerce faster than Western markets.
McKinsey’s Six-Level Automation Curve
McKinsey’s companion report, “The Automation Curve in Agentic Commerce” (January 2026), provides the most comprehensive framework for understanding where different types of agentic commerce fall on the maturity spectrum. A critical insight from the framework: this is a curve, not a ladder. Higher levels are not always better. The goal is what McKinsey calls “optimal delegation” – matching the right level of autonomy to the right type of purchase.
Level 0 – Programmatic Convenience. Rule-based automation for recurring purchases. Example: Amazon Subscribe & Save. No AI reasoning involved.
Level 1 – Assist. AI supports research and analysis without executing any actions. The agent finds options and compares specifications. The human makes all decisions and completes all transactions.
Level 2 – Assemble. The qualitative turning point. Agents begin orchestrating: resolving trade-offs, handling taxes and shipping calculations, and returning checkout-ready carts. The consumer’s role shifts from comparing options to approving proposed solutions.
Level 3 – Authorize. Consumers delegate rule-based purchasing. The agent executes transactions when predefined conditions are met. Example: “If my preferred running shoes drop below $80 from a trusted merchant, buy them.” The agent escalates only when conditions fall outside parameters.
Level 4 – Autonomize. Agents operate toward long-term objectives, not individual transactions. They anticipate needs, compare across merchants over time, and continuously optimize spending and loyalty goals.
Level 5 – Network Autonomy. Still nascent. Multiple specialized agents negotiate with each other on behalf of consumers – personal agents bargaining with merchant agents over pricing, logistics agents coordinating delivery, payment agents handling authorization. Transactions are settled through shared protocols.
Most consumer-facing agentic commerce today operates between Levels 1 and 2. Leading implementations are pushing into Level 3. Levels 4 and 5 remain largely theoretical for consumer retail, though B2B procurement is advancing faster toward higher autonomy levels.
Willingness to delegate varies sharply by purchase type. Routine, low-stakes categories like groceries and household essentials are natural candidates for higher autonomy. Identity-driven categories like luxury goods and fashion tend to stall at Levels 1 and 2, where discovery and personal expression are part of the value.
What Merchants Need to Do
Based on guidance from McKinsey, BCG, and Forrester, merchants should focus on four high-level priorities.
1. Make your product data machine-readable. AI agents cannot buy what they cannot understand. This means clean, structured product catalogs with accurate descriptions, real-time inventory data, standardized attributes (using schema.org markup and GS1 standards), and high-quality images. Poor product data is already costing businesses an average of $15 million per year in lost revenue (Mirakl). In an agentic world, the penalty compounds – agents will simply skip merchants whose data they cannot parse.
2. Adopt emerging commerce protocols. Evaluate UCP, ACP, and related standards. Shopify merchants can enable Agentic Storefronts to be discoverable across AI surfaces today. The merchants who are machine-accessible first will capture disproportionate agent-driven traffic during the adoption ramp.
3. Build trust infrastructure. Support agent authentication, transparent pricing, clear return policies, and auditable transaction logs. Trust is the primary barrier to consumer adoption of autonomous purchasing. Merchants who make it easy for agents to verify and communicate trustworthiness will be favored by the algorithms.
4. Invest in identity resolution and real-time data. Agents operate across surfaces – ChatGPT today, Google Gemini tomorrow, a WhatsApp bot next week. Merchants need the ability to recognize and serve the same customer across all of these touchpoints, with consistent pricing, inventory, and personalization.
Key Statistics at a Glance
| Statistic | Value | Source |
|---|---|---|
| Consumers using AI in buying journey | 45% | IBM-NRF (Jan 2026) |
| Shopping-related AI query growth (2024–2025) | 4,700% | Adobe/BCG |
| AI traffic to US retail sites YoY growth (Black Friday 2025) | 805% | Adobe |
| Consumers comfortable completing purchase through AI | 17% | ChannelEngine |
| Trust AI to place orders autonomously | 14% | Clutch (Jan 2026) |
| Global agentic commerce projected by 2030 | $3–$5 trillion | McKinsey |
| US agentic commerce projected by 2030 | $300B–$500B | Bain & Company |
| B2B agent-mediated spending projected by 2028 | $15 trillion | Gartner |
| ChatGPT daily shopping queries | 53 million | Digital Commerce 360 |
| ChatGPT weekly active users | 800 million | OpenAI |
| Amazon Rufus conversion lift vs. non-AI sessions | 60% higher | Amazon |
| AI-engaged shoppers conversion rate vs. non-AI | 12.3% vs. 3.1% (~4x) | Industry data |
| Shopify merchants on OpenAI commerce | 1 million+ | Shopify |
| Retailers exploring or implementing AI agents | 96% | BCG |
| Organizations planning agentic AI deployment | 99% | KPMG |
| Enterprises with agents in production (2025) | 52% | Google Cloud |
| Investment in agentic AI startups since 2023 | $9.7 billion | Market.us |
| Cyber Week 2025 orders influenced by AI agents | 20% ($67B) | Salesforce |
| Gen Z using AI for product research | 33% | Forrester |
| Boomer AI adoption growth (YoY) | 92% | IBM-NRF |
Frequently Asked Questions
How is agentic commerce different from e-commerce? Traditional e-commerce requires humans to search, compare, select, and checkout manually. Agentic commerce delegates some or all of those steps to an AI agent. The consumer expresses intent; the agent handles execution. Think of it as the difference between driving yourself to the store and sending a personal shopper who knows your preferences.
Is agentic commerce the same as conversational commerce? No. Conversational commerce refers to buying through messaging interfaces like WhatsApp, SMS, or live chat. Agentic commerce is broader – it includes any transaction where an AI agent acts on behalf of the consumer, whether through a chat interface, a voice assistant, a browser, or a background process. Conversational commerce can be a channel for agentic commerce, but agentic commerce also operates in non-conversational contexts.
What is the difference between UCP and ACP? UCP (Universal Commerce Protocol), created by Google and Shopify, is a decentralized standard where merchants publish their capabilities and any agent can discover and transact with them. ACP (Agentic Commerce Protocol), created by OpenAI and Stripe, is a platform-mediated standard where merchants submit product data to OpenAI and transactions happen inside ChatGPT. UCP is open and multi-agent. ACP is optimized for the ChatGPT ecosystem. Both are open-source, and many merchants will ultimately support both.
Can AI agents actually complete a purchase today? Yes. ChatGPT Instant Checkout (launched February 2026) allows users to buy products from over one million Shopify merchants directly inside the chat. Amazon Rufus assists with purchasing across Amazon’s catalog. Google AI Shopping Mode enables agent-assisted transactions. These are real, production systems processing real transactions. However, fully autonomous purchasing (where the agent buys without any human confirmation) is still limited to a small percentage of transactions.
What types of products are best suited for agentic commerce? Routine, specification-driven purchases are the most natural fit: groceries, household essentials, basic consumables, and replenishment items. Complex but research-heavy categories like electronics and travel also benefit significantly from agent-assisted research and comparison. Identity-driven categories like luxury goods and fashion are slower to adopt because the discovery and selection process is part of the value for the consumer.
Will agentic commerce replace my existing online store? No, but it will add a new channel. Your website, app, and marketplace listings will continue to matter. However, an increasing share of traffic and transactions will originate from AI agents rather than human browsers. The merchants who make their stores machine-readable and protocol-compatible will capture this new traffic. Those who do not risk becoming invisible to a growing segment of buyers.
How do AI agents handle payment security? The emerging protocol stack addresses this through multiple mechanisms. AP2 uses Verifiable Digital Credentials (VDCs) – cryptographically signed authorizations that specify exactly what an agent is permitted to purchase and under what conditions. ACP uses Shared Payment Tokens through Stripe, which are scoped to specific amounts and sellers. Visa’s Intelligent Commerce platform provides additional authentication layers. These systems are designed so that agents never handle raw payment credentials.
What should my business do first? Start with data. Audit your product catalog for completeness, accuracy, and machine readability. Ensure you have structured data markup (schema.org), accurate real-time inventory, and clear product descriptions. If you are on Shopify, enable Agentic Storefronts. Then evaluate UCP and ACP integration based on your platform and customer base. The merchants who move first on data quality and protocol adoption will have a structural advantage as agent traffic scales.
How fast is consumer adoption growing? Extremely fast. AI application usage surged 62% over two years (IBM-NRF). GenAI traffic to retail sites grew 4,700% year-over-year. Forty-eight percent of consumers planned to use AI during year-end 2025 sales, up nine percentage points from the prior year (BCG). And the fastest-growing demographics are older consumers – Boomer adoption grew 92% and Gen X grew 82% – suggesting this is not a niche trend limited to younger audiences.
What happens to SEO in an agentic commerce world? Traditional SEO remains important but is no longer sufficient. Merchants must also optimize for what the industry calls Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO) – ensuring that AI models can parse, understand, and cite your product information when generating answers. Gartner predicts a 25% drop in traditional search engine volume by 2026. The new imperative is not just ranking on a results page but being selected by an AI agent on behalf of the consumer.
Glossary of Key Terms
Agentic Commerce. A model in which AI agents research, negotiate, and complete purchases on behalf of consumers or businesses.
AI Shopping Agent. An AI system that discovers, evaluates, and transacts autonomously across the full shopping journey – from understanding intent to executing payment.
UCP (Universal Commerce Protocol). An open-source standard by Google and Shopify for agent-to-merchant discovery and transactions.
ACP (Agentic Commerce Protocol). An open standard by OpenAI and Stripe enabling programmatic commerce within ChatGPT.
AP2 (Agent Payments Protocol). A Google-led open protocol for secure agent-initiated payments using Verifiable Digital Credentials.
MCP (Model Context Protocol). An Anthropic-developed protocol standardizing how AI agents connect to external tools and data sources.
A2A (Agent-to-Agent Protocol). A Google-created open protocol enabling communication between multiple AI agents.
GEO (Generative Engine Optimization). The practice of optimizing content to be cited by AI-powered search and answer engines.
AEO (Answer Engine Optimization). Optimizing for placement in AI-generated answers; largely synonymous with GEO.
Automation Curve. McKinsey’s six-level framework (Level 0 through Level 5) for classifying the maturity of agentic commerce implementations, from rule-based convenience to multi-agent network autonomy.
Verifiable Digital Credential (VDC). A cryptographically signed digital object used in AP2 to authorize and contextualize agent-initiated transactions.
Instant Checkout. An ACP feature allowing ChatGPT users to purchase products directly inside the chat interface.
Agentic Storefronts. A Shopify feature enabling merchants to publish product data once and surface it across all major AI agent platforms.
Sources: McKinsey & Company (Oct 2025, Jan 2026), IBM-NRF (Jan 2026), Bain & Company, Morgan Stanley (Dec 2025), Gartner (Nov 2025), Forrester (Dec 2025), BCG (2025–2026), Adobe, Salesforce, ChannelEngine, Clutch (Jan 2026), eMarketer (Dec 2025), Google Cloud, KPMG, Digital Commerce 360, Precedence Research, Market.us, Wipro, Mordor Intelligence, Capital One Shopping, Capgemini, Mirakl.
Hexagon Team
Published March 8, 2026
