What Agentic Commerce Actually Is (and Isn't)

It shows up in conversations about AI-powered search, shopping assistants, product recommendations, autonomous purchasing, and new ways to complete checkout. But what does agentic commerce actually mean?

McKinsey describes it as "shopping powered by intelligent AI agents capable of anticipating, personalizing, and automating every step of the process."^[1]

But what does that look like in practice? Is it asking ChatGPT to find the best running shoes for your budget? Is it a retailer's chatbot recommending three products from its own catalog? Is it an AI assistant comparing prices and availability across several retailers? Or is it an agent that selects the product, places the order, and schedules delivery without the shopper returning to make the final decision?

Those experiences may look increasingly similar to the shopper, but they represent very different levels of autonomy. The more broadly the term is used, the less clear it becomes where AI-assisted shopping ends and truly agentic commerce begins.

That distinction matters when retailers are being asked to place technology bets, rethink the path to purchase, and prepare for a market that some forecasts put in the hundreds of billions of dollars.^[1][2] So what is agentic commerce, really—and what isn't it?

The Difference Is Delegation

Agentic commerce is not a smarter chatbot. It is the delegation of a purchase decision.

That means not every use of AI in shopping is agentic commerce. A chatbot answering a product question isn't agentic. Neither is a recommendation engine ranking products from one retailer's catalog. Even AI search that researches and compares products may still leave the decision and action entirely with the shopper.

For retailers trying to separate AI assistance from agency, it helps to look at three things:

• The goal: Is the AI working toward an outcome the shopper has defined rather than simply answering a question?
• The consideration set: Is it independently finding, evaluating, and narrowing the group of products the shopper is actually considering?
• The action: What has the shopper authorized the AI to do—from recommending the next step to completing a transaction within defined limits?

Autonomous checkout is one end of the spectrum, not the threshold for agentic commerce. The defining shift is delegation: the shopper gives an AI system greater responsibility to work toward an outcome and act on their behalf.

As shoppers delegate more of the journey, retailers must prepare for AI systems to become a new intermediary between their products and their customers.

The Delegation Gap

Consumers are already bringing AI into the shopping journey.

Adobe reported in April 2026 that 39% of surveyed U.S. consumers had used AI for online shopping,^[3] while Bain estimates 30–45% of U.S. consumers use generative AI for product research and comparison.^[2] Adobe also found AI-referred traffic—visits arriving at retail sites from AI tools—rose 393% year over year in Q1 2026. In March, that traffic converted 42% better than non-AI traffic, compared with 38% worse a year earlier.^[3]

But using AI to shop is different from letting AI decide.

In a May 2026 Gartner survey, no more than 11% of U.S. consumers said they would let AI make a purchase decision, even in lower-stakes categories. And 54% of surveyed consumers who had used generative AI said they had needed to double-check the information it provided.^[4]

That is the Delegation Gap: AI is becoming part of how consumers shop faster than consumers are becoming willing to let AI shop for them.

The gap also helps explain why estimates of the opportunity vary so widely. McKinsey estimates that agentic commerce could orchestrate $900 billion to $1 trillion in U.S. B2C retail revenue by 2030.^[1] Bain estimates $300 billion to $500 billion, or 15% to 25% of U.S. e-commerce.^[2] One important difference: Bain excludes journeys that use AI only for search or discovery.

How you define agentic commerce can change the size of the market by hundreds of billions of dollars.

AI Doesn't Need the Buy Button to Matter

The Delegation Gap doesn't give retailers permission to wait. AI can influence a purchase long before a consumer delegates the final decision.

Similarweb data reported by Search Engine Land shows that for searches for "headphones," the share displaying AI Overviews rose from 2.28% to 32.76% between January 2025 and January 2026, while organic click share fell from 73% to 50% over the same period.^[5] The data does not establish that AI Overviews caused the decline, but it illustrates how product discovery is changing.

It also points to a broader shift toward the Zero-Click Journey: discovery, comparison, and potentially even the transaction can happen inside an AI interface without the shopper visiting the retailer's site. For retailers, that makes the consideration set critical.

If an AI system cannot find, understand, or trust a product, that product may never make it far enough to be compared.

Most Agent-Readiness Isn't AI Work

For retailers, the real challenge is architectural.

Adobe found that the average U.S. retail product detail page was only 66% machine-readable—meaning its product information could be reliably identified and interpreted by machines.^[3]

For many retailers, the next investment in agentic commerce shouldn't be an agent at all. It should be the infrastructure agents depend on:

• Product truth: Structured, machine-readable product data with complete attributes and product identifiers such as Global Trade Item Numbers (GTINs).
• Price and availability: Direct, reliable access to current pricing and inventory.
• Access: Clean feeds and application programming interfaces (APIs) that let other systems retrieve that information.
• Governance: Clear ownership for keeping product information accurate and accessible.

Known facts shouldn't need to be inferred by AI. They should come from systems built to provide reliable answers.

Model Context Protocol (MCP), an open standard for connecting AI systems to external data and tools, and other emerging standards may change how agents access that information. But no single protocol should dictate the strategy. Retailers need trustworthy product data that can be exposed through whichever standards emerge.

Prepare for Delegation, Not the Buzzword

Agentic commerce may still be evolving, but the foundations retailers need are already clear.

Retailers don't need to predict when consumers will hand over the Buy button to prepare for greater delegation. They need to make sure their products can be found, understood, compared, and acted on as AI takes on more of the shopping journey.

The goal isn't to become more "agentic." It is to be ready when your customers are.

Key Action Items

Audit what agents can read. Identify gaps in product pages, structured data, attributes, identifiers, and feeds.

• Expose price and inventory reliably. Give systems direct access to accurate availability and pricing through maintained feeds or APIs.
• Measure outcomes, not autonomy. Track AI-referred traffic, conversion, revenue per visit, and catalog readability—not how many agents or protocols you support.
• Assign ownership of product truth. Establish accountability across merchandising, e-commerce, and technology for the information AI systems rely on.