The state of agentic commerce

Anywhere from about a quarter to three quarters of them, depending on whose survey you read. Adobe puts it at 39% of US consumers, IBM and the NRF at 45% globally, and Riskified at 73%.

That spread looks like disagreement, but most of it comes from what each survey counted. Morgan Stanley asked who had bought something using AI in the past month and got 23%. Riskified asked who uses AI anywhere in their shopping journey and got 73%. Most studies land in between because they ask about using AI tools to shop, which covers everything from a single product question to a full comparison. When you see a single adoption number quoted in a deck, the first thing to ask is what it measured.

The broader studies also show where AI enters the journey. Adobe found that 55% of consumers turn to AI for inspiration and ideas, most often before they begin shopping, and IBM and the NRF found 41% use it to research products. For your team, that means a shopper may arrive with a shortlist built somewhere you couldn't see, and the first conversation they have with your brand is about narrowing it down.

Mostly no, and the reason is control. In Accenture's 2026 survey of 25,590 consumers in 16 countries, 32% would let an agent decide what to buy as long as they make the payment themselves and 9% were open to fully autonomous shopping. In Spreedly's US survey, 64% would not allow AI to spend without their approval.

Payment is where the hesitation gets sharpest. Visa found that only 23% of US consumers trust generative AI to handle payment transactions for them. McKinsey's European research, published under the title AI to browse but not to buy, found 56% of consumers are comfortable with AI suggesting options as long as they keep the final say.

Comfort is growing, though. Adyen measured AI assistant use among US shoppers jumping from 12% to 35% in a single year, and comfort with delegation tends to follow familiarity. For now, though, the shopper who wants AI's help and still wants to press the button is the one you're designing for, and an experience that assumes they'll hand over the whole purchase is designing for about one shopper in ten.

It converts better than the rest of your traffic now, which wasn't true a year ago. Adobe has tracked the same comparison across several reports. During Prime Day 2025, shoppers who arrived from AI tools converted 23% worse than paid search, email, and social traffic. By the 2025 holiday season they converted 31% better than other traffic sources, and in May 2026 they converted 54% better and generated 53% more revenue per visit. And during Prime Day 2026, AI traffic converted 40% better than non-AI channels.

The likely reason is the research that happens before the click. A shopper who has already compared options with an assistant shows up closer to a decision. Adobe's survey adds a post-purchase angle too: 69% of consumers said they're less likely to return an item they bought with an AI assistant's help.

The volume is still small. Bain found AI accounts for up to a quarter of referral traffic for some retailers, and still less than 1% of total traffic. So AI-referred visitors are worth treating as their own segment, with their own conversion and revenue numbers, but they won't move your top line on their own this year. Their behavior is a preview of how more of your shoppers will arrive.

Shoppers trust the retailer's own agent most. Bain's research found shoppers trust a retailer's on-site agent three times more than a third-party agent to complete transactions, and a likely reason is accountability. If an assistant on your site gets a size wrong or misses a return window, the shopper knows who answers for it.

That matters because so much of the agentic commerce conversation is about third-party agents shopping across the web for someone. Those agents may get there, but the trust today sits with the brand whose site the shopper is already on. Adobe's 2026 holiday forecast found 77% of shoppers who've used AI for online shopping feel more confident in their purchase, and confidence is the thing a retailer's own assistant is best placed to give, because it can see the catalog, the stock and the policies and answer for all of them.

Crate & Barrel's point applies well beyond AI. A shopper in a conversation on your site has already shown intent, and helping them buy is the job.

Most aren't, though most plan to be. In TCS's 2026 retail study, 85% of retailers weren't yet using multi-agent AI systems and 48% had no plans to. IBM and the NRF found 51% of executives cite limited AI expertise as a barrier. Meanwhile Deloitte found nearly 68% of retail executives expect to deploy agentic AI in the next 12 to 24 months.

The foundation work is further behind than the plans. In PayPal's survey of US merchants testing agentic tools, only about one in five had 80% or more of their product catalog available as structured, machine-readable data. An agent, yours or anyone else's, can only recommend what it can read.

The forecasts explain why everyone is planning anyway: Morgan Stanley projects US agentic commerce sales of $190 billion to $385 billion by 2030 (base to bull case), while Juniper Research, looking at the global market, describes the market as only pilot deployments in 2025 and 2026. The money is still a projection and the pilots are what exist today, so the retailers who get their catalog and on-site experience in shape during the pilot years will be ready when the volume shows up.

Use the scorecard on the left to see where you stand against the market.

Build now. Start with catalog data an agent can read, since most merchants still don't have 80% of theirs in that form. Then put AI on your own site that knows the customer, because that's where shoppers say their trust is. Both pay off whether third-party agents take off next year or in five years.

Measure. Separate AI-referred traffic in your analytics and track it by conversion rate and revenue per visit. Session counts will undersell it, since the volume is small and the intent is high, and you'll want your own numbers next to Adobe's before you make a bigger bet.

Hold off. Heavy investment in fully autonomous checkout through third-party agents can wait until shoppers catch up. Five percent of US consumers have used one, and fewer than one in four US consumers trust AI with payment. Watch the protocols and payment rails as they mature, and keep your catalog ready to plug in when they do.

The data points one way. Shoppers want AI's help, they want the final say, and they trust a brand's own site more than a third-party agent. That's the gap Gladly is built for.

Gladly brings the AI conversation onto the site you own. It helps shoppers find, compare and add products to their cart, and can check them out inside the conversation, with the shopper making the call. It knows the customer, since every past chat, call and email sits on one record, so recommendations reflect what they've bought and asked about before. When a shopper wants a person, the conversation moves to your team with full context and nobody starts over. The same conversation handles service after the sale, where trust is kept or lost.

At Balsam Hill, nearly one in five shoppers who viewed a product in chat added one to their cart, and chat drove six figures in add-to-cart value in the first 30 days.