What Agentic Commerce Means and Why Its Market Forecasts Disagree
A plain agentic commerce definition, how McKinsey, Bain, Gartner and the protocol authors draw its boundary, and why the 2030 market size forecasts differ.
On this page
- What does agentic commerce mean in plain terms?
- Where each source draws the boundary
- What does Gartner say about agentic commerce?
- How big do the forecasts say it will be?
- Why do the numbers differ so much?
- How much ecommerce runs through AI agents in 2026?
- Quoting a forecast so it survives questions
- Sources
Agentic commerce is shopping in which an AI agent does some or all of the buying work for a person or a business, from finding and comparing products to paying for them. The sources quoted below agree on that much and part ways on the boundary: whether a purchase the agent only influenced counts, whether the agent has to complete checkout, and whether business buying is in scope. That boundary, along with the market each forecast measures, helps explain the distance between a 2030 forecast of $190 billion and one of $1 trillion for the same country. We read the primary pages behind the best-known definitions and forecasts on October 7, 2026.
What does agentic commerce mean in plain terms?
The agentic commerce definition that the main sources share is short: software agents act for a buyer across the steps of a purchase, while the buyer sets the intent and the limits. McKinsey calls it "shopping powered by AI agents acting on our behalf" in an October 2025 report, and describes agents that anticipate consumer needs, navigate shopping options, negotiate deals and execute transactions through multistep chains of actions. The Stripe documentation states it from the payments side: "Agentic commerce uses AI agents to support transactions between buyers and sellers, and to let agents transact on behalf of the people they serve."
Read together, the two describe a chain that runs from discovery and comparison through cart, checkout and payment, and a definition is mostly a decision about which links to count. The first links are familiar ground for teams working on generative engine optimization, since an agent has to find and trust a product before it can buy it.
Where each source draws the boundary
McKinsey draws the widest line. Its report covers the full journey and names three interaction models: agent to site, agent to agent, and brokered agent to site, where an intermediary system sits between a shopper's agent and a platform.
Bain counts influence and stops short of search. Its December 2025 note says: "We define agentic commerce as purchases that are initiated, influenced, or completed by third-party AI agents or retailer-hosted agents, excluding shopping journeys that only use AI-assisted search or discovery." Under that wording a retailer's own on-site agent is in, and a shopper who only asked a chatbot for ideas is out.
Morgan Stanley Research writes about "agentic shoppers", AI-powered assistants that scout for hard-to-find items, compare prices in real time and anticipate recurring purchases such as groceries with minimal user intervention.
The protocol authors start at the transaction. The Agentic Commerce Protocol site, developed by Stripe and OpenAI, opens with this line: "An open standard for programmatic commerce flows between buyers, AI agents, and businesses." Its scope is coordinating checkout and passing payment credentials securely, with the business staying merchant of record. The Agent Payments Protocol documentation, which carries a Google copyright and says standardization will continue in FIDO working groups, starts from the observation that payment systems assume a human is clicking buy on a trusted website, and it separates human-present from human-not-present transactions. We found that split in none of the analyst pages, and it matters for sizing: purchases made with no human present are a subset of all agent-assisted orders.
What does Gartner say about agentic commerce?
Gartner's public statements on agentic commerce arrive as predictions and survey findings, and we found no free Gartner page that defines the term in a single sentence.
A widely repeated number comes from its October 2025 strategic predictions. The Gartner release states: "By 2028, 90% of B2B buying will be AI agent intermediated, pushing over $15 trillion of B2B spend through AI agent exchanges". The prediction concerns business buying. Under a separate prediction on programmable money, the release writes of "machine customers, such as AI agents with economic agency". We did not find the phrase agentic commerce in that release.
Gartner's consumer research is cooler. A May 2026 Gartner release, which does use the term, reports a January 2026 survey of 322 US consumers in which willingness to let AI make purchase decisions topped out at 11%, in lower-stakes categories such as personal care and household supplies. Narrowing choices was more acceptable: 31% would allow it for household supplies and 28% for personal electronics. The release advises marketers to favor research, comparison and deal-finding tools over fully autonomous shopping agents.
How big do the forecasts say it will be?
The agentic commerce market size and growth forecasts we compare here in 2026 all look past this year. Three point at 2030 and one at 2028, and we found no dollar estimate for 2026 itself on any of the four pages.
| Source and date | Headline figure | Scope | What is counted | Stated basis |
|---|---|---|---|---|
| McKinsey, October 17, 2025 | $900 billion to $1 trillion in the US and $3 trillion to $5 trillion globally, by 2030 | B2C retail, goods only | Revenue agents could orchestrate | Moderate merchant-readiness assumptions, US Census data, forecasted adoption rates |
| Bain, December 17, 2025 | $300 billion to $500 billion by 2030, roughly 15% to 25% of e-commerce | US e-commerce | Purchases initiated, influenced or completed by agents | We found no method note on the page |
| Morgan Stanley Research, December 8, 2025 | $190 billion to $385 billion by 2030, a 10% to 20% share | US e-commerce spending | Spending by agentic shoppers | 10% is the likely case, 20% the optimistic one; we found no model detail |
| Gartner, October 21, 2025 | Over $15 trillion by 2028, 90% of B2B buying | B2B buying, no geography given | Spend through AI agent exchanges | A strategic prediction; the full research is for Gartner clients |
Why do the numbers differ so much?
Three things differ between these forecasts: the pool, the counting rule and the buyer. Take the pool first. Bain and Morgan Stanley both express the forecast as a share of US e-commerce, so dividing each dollar figure by its share gives the base each implies. By our arithmetic on their published figures, Morgan Stanley's $190 billion at 10% and $385 billion at 20% both imply about $1.9 trillion of US e-commerce in 2030, and Bain's $300 billion at 15% and $500 billion at 25% both imply about $2 trillion. The two start from nearly the same market and disagree on the share. Bain's definition explicitly counts purchases an agent influenced. We found no statement in Morgan Stanley's public article on whether those count, so a reader cannot tell how much of the gap is definition and how much is a different view of adoption.
McKinsey's US number is larger than both, and it measures something else. Its stated pool is the US B2C retail market, a broader label than e-commerce, and its unit is revenue that agents could orchestrate, a term we found no definition for on the page. It adds its own caveat: "These figures only reflect goods and do not yet include services; nor do they account for the significant B2B marketplace."
Gartner's $15 trillion belongs to a different market: business buying, two years earlier, counted as spend passing through agent exchanges. It should never share an axis with the consumer figures.
A hypothetical shows what the spread does to a plan. Assume a US retailer expects $40 million of online sales in 2030 and applies each market-wide share to itself, a shortcut we found on none of these pages. Morgan Stanley's 10% to 20% gives $4 million to $8 million flowing through agents, and Bain's 15% to 25% gives $6 million to $10 million. The low and high ends differ by a factor of 2.5 before any category difference is considered.
How much ecommerce runs through AI agents in 2026?
Agentic commerce market growth in 2026 is easier to see in ecommerce shoppers' use of AI agents for research than in completed purchases. We found no measured share of 2026 sales completed by agents on the pages we read. The evidence there is survey data, and it puts research use ahead of delegated buying.
- Morgan Stanley's AlphaWise survey of US adoption, reported in December 2025, found that about half of large language model users had researched and compared product prices in the past month and that 30% to 40% reported making purchases. From adoption rates the bank estimates that roughly 23% of Americans bought something via AI in that month. We found no sample size in the public article.
- Bain's December 2025 note says 30% to 45% of US consumers use generative AI for product research and comparison, and that most are not yet comfortable letting AI handle an end-to-end transaction.
Set beside Gartner's 11% ceiling, the fair reading for 2026 is that discovery through assistants is common, purchase handoff is early, and every large number is a projection for 2028 or 2030. An agent can only buy what it can find, describe accurately and trust, so the measures behind AI share of voice and the practices for getting cited by ChatGPT, Perplexity and AI Overviews are sensible early indicators for agentic sales. The AI Visibility Checker is a further resource on the visibility side, and our AI search optimization page covers the service side.
Quoting a forecast so it survives questions
When one of these numbers goes into a deck, carry four labels with it: the year, the geography, the pool (retail, e-commerce or B2B spend) and the counting rule (influenced, orchestrated or completed). A slide that says up to $1 trillion by 2030 and one that says $190 billion by 2030 are both accurate citations, and they describe different things. The same habit applies to visibility statistics, which we cover in what the published AI citation numbers can support. A forecast quoted with its scope attached tells the audience what the forecaster was measuring.
Sources
- McKinsey: The agentic commerce opportunity, October 2025
- Bain: 2030 Forecast snap chart, December 2025
- Morgan Stanley: Here Come the Shopping Bots, December 2025
- Gartner: strategic predictions release, October 2025
- Gartner: AI shopping survey release, May 2026
- Stripe: Agentic commerce documentation
- Agentic Commerce Protocol: home page
- Agent Payments Protocol: AP2 documentation