Three in four holiday shoppers in the United States, the United Kingdom, Canada and Australia are not interested in having a personal AI agent make purchases for them, according to survey results Salesforce published today. The same article projects that AI will drive $75.8 billion in Cyber Week spending this year, up 13% from 2025, without explaining how that figure was calculated or what counts as AI-driven.

The findings appear in a Salesforce newsroom article titled "New Data: Holiday Shoppers Will Rely on Agents. Retailers Must Be Ready." They draw a line between two uses of the technology. Shoppers ask AI what to buy. Far fewer let it pay. Only 19% of people who use a personal AI agent have deployed it at checkout, according to Salesforce, and just 9% give their agent unrestricted access to their financial details.

In Short

Salesforce surveyed more than 2,500 shoppers in four countries and found that most use AI to compare prices and read reviews, but 76% do not want an AI agent buying things for them. That matters to retailers and advertisers because almost 20% of shoppers now say AI is their first stop, which moves part of product discovery away from retailers' own websites. What changes is the pressure on product data, prices and how sites treat automated visitors, although several of Salesforce's strongest sales claims for retailer-run agents come without a published method.

Two kinds of data in one article

The article combines two sources, and they are not described in the same detail.

The first is a Salesforce Consumer Shopping survey of "more than 2,500 shoppers in the U.S., the U.K., Canada, and Australia." Every chart on the page credits that survey. The article does not give fieldwork dates, the number of respondents per country, whether results were weighted, or a margin of error. It also does not break any figure out by country, so a reader cannot tell whether US respondents behave differently from Australian ones.

One detail narrows the timing. Meta released its Muse agent in the US on September 8, 2026, and the survey asked about Muse use, which means at least that question was put to respondents after that date. Muse's buying agent launched for US users, and the article does not say how the 12% usage figure splits across the four countries.

The second source is described only as "Salesforce proprietary research." Three figures come from it: AI is projected to drive $75.8 billion in Cyber Week spending, up 13% from 2025; AI-driven recommendations are expected to influence more than one in five holiday orders; and traffic flowing to ecommerce sites from AI tools has "more than doubled, up 104% year over year." The article does not say whether the $75.8 billion is a global or a national figure, which days count as Cyber Week, what qualifies as AI-driven spending, or which period the 104% traffic growth covers. Taken at face value, a 13% rise to $75.8 billion implies a 2025 base of about $67.1 billion, according to a PPC Land calculation.

The distinction matters because the survey measures what people say they do, while the projection and the traffic figure appear to rest on transaction and site data that the article does not describe. Salesforce presents them side by side.

AI as the first stop

Almost 20% of shoppers now say that AI is their first stop on the way to the store, according to the survey, and 74% trust the recommendations it gives them. The sentence does not make clear whether the 74% refers to all respondents or only to those who use AI tools. A separate Salesforce post on 2026 holiday predictions, linked from the same article, phrases the same 74% as trust in product recommendations received from AI chat.

The tools are also changing which brands get seen. Thirty percent of what Salesforce calls "AI shoppers" say these tools frequently introduce them to brands they did not know. The article does not define an AI shopper or say how many respondents fall into that group. The influence runs in both directions: 11% of shoppers say they would skip a purchase if an AI assistant advised against it.

"AI is no longer a niche tool just for tech-savvy shoppers," said Caila Schwartz, Head of Agentic Commerce Shopper Insights at Salesforce. "Everyday shoppers are asking AI what to buy, where to buy it, and whether it's worth the money before they ever land on a retailer's site."

The article opens with a hypothetical Cyber Monday shopper who asks AI to find the best price on a stand mixer, check the reviews and flag anything that will not arrive before December 20, and then closes the app and buys it themselves. That scenario sums up the survey's central finding better than any single number in it.

Muse leads a young field of personal agents

Thirty-one percent of consumers have tried a personal AI agent, according to Salesforce. Meta Muse is the most widely adopted, used by 12% of consumers surveyed, and it is also the most recognized, followed by Grok Bot, the agent product from xAI. If both shares use all respondents as their base, which the article implies but does not state, close to four in ten personal agent users have used Muse, according to a PPC Land calculation.

Among people who use them, agents are becoming a daily habit: 28% of personal agent users give their agent tasks multiple times a day. The most common shopping tasks are price comparison, analyzing product specifications and summarizing reviews for "sizing quirks or technical differences," and broad discovery, such as brainstorming gifts against a set of criteria. None of those tasks requires the agent to buy anything.

Muse's lead in the survey matches what traffic measurement firms have recorded since its release. HUMAN Security put Muse at 72% of daily AI agent requests in its early observations at the end of September, and DataDome found Muse accounted for 33% of agentic browser requests in September despite only nine days of measurement in the month. Not every retailer welcomes it. On September 20, Amazon said it had blocked Muse from shopping on Amazon.com, saying the agent did not identify itself when browsing and appeared to capture and store customer credentials.

Checkout is where shoppers stop

The survey finds a sharp drop between research and payment. Only 19% of personal agent users have deployed their agents at checkout. If the 31% and 19% figures share the same base, roughly 6% of all respondents have let an agent take part in a purchase, according to a PPC Land calculation.

Those who do are cautious. Just 9% of personal agent users give their agents unrestricted access to their financial details, and 36% have set up a separate credit card just for their agent. The article introduces these two figures as limits set by "shoppers who do let agents buy," but states both as shares of all personal agent users, so the share among actual agent buyers is not given.

About three-quarters of shoppers, 76%, are not interested in having a personal AI agent make purchases for them at all. A chart in the article presents the same split as 76% "not comfortable" and 24% "comfortable" having an agent make purchases on their behalf. The top barriers, in order, are unwanted sharing of purchase history or personal data, accidental or unwanted charges, and buying from a fraudulent business. The third barrier appears only in the chart.

Retailers and platforms have started to build the kind of guardrails that some shoppers are improvising with a second card. Muse asks for approval before purchases and pays with a one-time card through Link by Stripe. Walmart gives each agent its own profile, with approval for every purchase by default or a customer-set spending limit such as $50, $100 or $200, and agents never see card numbers.

Other surveys have produced very different readings of shopper willingness, which says as much about question wording and samples as about consumers. A November 2025 study by PSE Consulting of 2,000 UK online shoppers found that 85% of those planning AI-assisted Christmas shopping would trust AI to place orders and execute payments. Research from Boston Consulting Group and Moloco put the share of consumers comfortable letting AI make purchases on their behalf at 45%. Salesforce's figure, 24%, is the lowest of the three.

Two or three mistakes

Trust, in Salesforce's framing, is fragile. Most shoppers will abandon an AI assistant after just two or three mistakes, according to the article, which gives no percentage for "most." When AI makes a bad recommendation, 32% of shoppers blame the AI. The article does not say whom the other 68% blame. The retailer? The brand? Themselves?

Errors are common. Among personal AI agent users, 27% say their agent has bought something that was too expensive, 22% have had their agent blocked from completing a purchase, and only 32% report no major errors at all. That leaves about 68% of agent users reporting at least one major error, according to a PPC Land calculation.

The tolerance threshold in Salesforce's data is looser than in some earlier research. A YouGov poll of 2,080 UK adults for ACI Worldwide, fielded in June, found that 60% would stop using an AI shopping agent after one mistake. The 22% block rate also has a counterpart on the retailer side: an audit by Blend Commerce found that bot protection refused its shopping agent at 29 of 200 retailers.

The retailer-side claims

The second half of the article turns from personal agents, which shoppers bring with them, to shopper agents that retailers run on their own websites. Salesforce sells such agents to retailers, and the article links the term to its own commerce product page.

Here the numbers grow larger. Shoppers who arrive through AI referrals are more than two times as likely to convert as typical ecommerce shoppers, according to Salesforce. Thirty-nine percent of shoppers have used a brand's shopper agent on an ecommerce site; 80% of them reported a good experience, and 75% of that group felt more loyal to the brand afterward. Applied to the whole sample, that chain means about 23% of shoppers both used a brand agent and came away feeling more loyal, according to a PPC Land calculation.

Shoppers who engage with a brand shopper agent spend more, with an 8% higher average order value, and buy 23% more items than the average order. And brands that deploy AI shopping agents "see seven times faster sales growth than brands without them," according to the article. The accompanying chart labels it "7x faster online sales growth."

A sevenfold figure without a method

Each of these charts carries the line "Source: Salesforce Consumer Shopping Survey." A consumer survey can capture whether a respondent used a brand agent and how they felt about it. Order values, item counts, conversion rate comparisons and brand-level sales growth are normally measured from transaction data, and the article does not explain how a survey of shoppers produced them, over what period, or how many brands sit in each group.

The sevenfold claim also sits awkwardly beside Salesforce's own earlier framing. The holiday predictions post linked from today's article says retailers using branded AI shopper agents saw 59% higher holiday sales growth in 2025, with sales up 6.2% for brands with shopper agents against 3.9% for brands without. That is a ratio of about 1.6 to one, not seven to one. The two claims may cover different periods or definitions, but the article does not reconcile them.

Other retailers have published comparable claims of their own. Amazon chief executive Andy Jassy said on October 29 that US customers using Alexa for Shopping spend 50% more per order than those who do not. Both figures compare users with non-users, and neither separates the effect of the agent from the kind of customer who chooses to use one. Microsoft Clarity, measuring behavior rather than spending, found checkout pages in 25.21% of sessions involving AI agents against 14.65% of human-only sessions.

Traffic that does not say what it is

The article closes on operations rather than marketing. As more shoppers send personal agents to compare prices and read reviews, retailers "will need to handle that nonhuman traffic without slowing their sites or degrading the experience for human shoppers," according to Salesforce, and their own shopper agents and product data "will also need to be ready for the questions those personal agents ask."

"My advice before Cyber Week is simple: Be the brand that AI can find and humans can trust," Schwartz said. "Shoppers are starting with AI this season, so make sure your product data is accurate, your prices and inventory are current, and your own agent is ready to answer the hard questions."

Handling that traffic is harder when most of it is anonymous. Arkose Labs found that 97.7% of the legitimate AI agent sessions it observed carried no verified identity, with only 2.3% presenting a Web Bot Auth signature. A retailer that cannot tell a shopper's agent from a scraper or a fraud tool faces a choice between blocking traffic that may be a sale and admitting traffic that may be an attack. The survey's 22% block rate does not say who stopped the agent: a retailer's security system, a payment provider or the agent's own limits.

The traffic numbers from different vendors are also hard to line up. Adobe, using US retail site data, forecast in September that AI traffic to US retail sites would rise 130% over the holiday season, with peaks of 159% on Thanksgiving, 95% on Black Friday and 88% on Cyber Monday, within a projected $275.1 billion in US online spending from November 1 to December 31. Salesforce's 104% describes traffic that has already doubled, with no stated period or market. The two figures measure different things and cannot be compared directly.

Salesforce has made the personal agent argument before. On October 19, a Salesforce executive argued that personal AI agents would add contact center traffic rather than reduce it, without supporting data. Today's article supplies survey data on shopping behaviour, but it does not address customer service volumes.

Why this matters for marketers

The survey describes a split that affects where marketing budgets can reach a shopper. Discovery is moving into AI tools: almost one in five respondents start there, 30% of AI shoppers say the tools introduce them to unfamiliar brands, and 11% would drop a purchase on an assistant's advice. Those moments happen inside AI tools rather than on a retailer's site, and the survey does not ask whether respondents saw sponsored placements while using them.

The purchase, by contrast, is still mostly human. With 76% of respondents unwilling to let agents buy and only 19% of agent users having done so, the visit that ends in a sale is likely to arrive as a referral from an AI tool rather than as an autonomous transaction. That shifts attention to what the agent reads before the click: product specifications, reviews, prices and stock levels, the same tasks respondents say they hand to agents most often.

The article is also a vendor document. Salesforce sells the shopper agents whose results it reports, and its largest figures, the sevenfold sales growth, the doubled conversion and the $75.8 billion projection, are the ones with the least published method. The survey numbers on shopper attitudes are more modest and more consistent with other research, and they point the same way: shoppers are using AI to decide, and keeping the card in their own hands for now.

Timeline

Summary

Who: Salesforce, through a newsroom article quoting Caila Schwartz, Head of Agentic Commerce Shopper Insights, and a survey of more than 2,500 shoppers in the US, UK, Canada and Australia. Meta's Muse and xAI's Grok Bot are named as the most recognized personal agents.

What: The survey found that 31% of consumers have tried a personal AI agent, 12% have used Muse, only 19% of agent users have deployed an agent at checkout, and 76% of shoppers do not want an agent buying for them. Salesforce also projected $75.8 billion in AI-driven Cyber Week spending, up 13% from 2025, and claimed that brands running their own shopper agents see 8% higher order values and seven times faster sales growth, without publishing a method for those figures.

When: Salesforce published the article today, November 12, 2026, two and a half weeks before Cyber Monday on November 30. The survey's fieldwork dates are not given.

Where: The survey covers the United States, the United Kingdom, Canada and Australia. The geography behind the $75.8 billion projection and the 104% traffic growth is not stated.

Why: Salesforce argues that retailers need product data, prices and their own agents ready for AI-led shopping. For marketers, the data indicates that discovery is moving into AI tools while the purchase remains largely human, which changes where influence on a sale is won and how it is measured.