Research

Salesforce forecasts $75.8B AI-shopping wave as checkout lags

Salesforce research associates AI agents with higher spending and loyalty, but only 19% of personal-agent users use them to check out. Consumer trust still trails technical ambition.

74% of shoppers trust AI agents for buying advice – but checkout is another story
74% of shoppers trust AI agents for buying advice – but checkout is another storyjustynafaliszek / pixabay
By James Calloway6 min read

Why it matters

  • Salesforce forecasts $75.8 billion in Cyber Week spending from AI-driven recommendations, up 13% from 2025.
  • A Salesforce survey of more than 2,500 shoppers found 74% trust AI recommendations, while 20% use agents for holiday research.
  • Agent-referral users were twice as likely to purchase, spent 8% more and bought 23% more items per order.
  • Only 19% of personal-agent users use agents to check out, while 73% of consumers are uncomfortable sharing payment information.
  • Traffic from AI tools to e-commerce sites increased 104% year over year.

Salesforce forecasts that AI-driven shopping recommendations will drive $75.8 billion in Cyber Week spending, up 13% from 2025. The forecast puts agent referrals at the center of holiday retail, but the same research shows a gap between accepting advice and authorizing purchases.

What did Salesforce measure?

Salesforce surveyed more than 2,500 shoppers in the US, UK, Canada, and Australia. ZDNET's summary says 20% of holiday shoppers worldwide use AI agents as their first stop for research and advice.

ZDNET also reports that traffic from AI tools to e-commerce sites has increased 104% year over year. In this research, AI agents handle product research, price comparisons, reviews, gift brainstorming and, for some users, checkout.

The source gives no survey dates, question wording, weighting details, country-level results or independent validation. Its worldwide framing also rests on a sample from four countries.

The traffic increase and spending forecast measure different things. The former tracks visits from AI tools, while the latter assigns future spending to AI recommendations. Without an attribution rule, a shopper who clicks an AI answer and later buys may sit between the two figures.

The research measures traction, not proof that agents caused incremental spending. The $75.8 billion forecast remains a forecast rather than an audited result.

Why do shoppers trust agents?

Salesforce found that 74% of shoppers trust recommendations from AI shopping agents. The trust carries consequences in both directions: 30% said an AI introduced them to a new brand, while 11% would avoid a brand if an agent told them to.

Salesforce also reported that 1 in 3 consumers has used an AI agent. It calls Meta Muse widely adopted and puts its usage at 12% of all consumers in the surveyed population.

Salesforce calls demand for Grok Bot high. ZDNET's summary gives no comparable usage figure for Grok Bot, so the two products' relative scale remains unmeasured.

The survey does not distinguish an agent embedded in a retailer's own site from a third-party assistant. It also does not disclose whether consumers use more than one service. Those omissions matter: Meta Muse's 12% and Grok Bot's undefined reach cannot be added together, and the broader trust percentages cannot be assigned to either product.

According to Salesforce, 28% of consumers give their agents multiple tasks per day. The most common uses are:

  • Comparing prices
  • Reading product reviews
  • Brainstorming gift ideas

Brand discovery also ranks as a major benefit of agent use, according to the report. Consumers have accepted agents as research assistants, but brand discovery remains a two-sided bet.

What changes when agents recommend a brand?

Shoppers who use AI agent referrals are twice as likely to make a purchase, according to Salesforce. The survey does not define the comparison group or establish causation, so the result may describe AI-oriented shoppers rather than an agent-created conversion lift.

Among the 39% of shoppers who had used an AI agent on an e-commerce site, 80% reported a positive experience. Salesforce also found that 75% felt more loyal to the brand after the agent experience.

The commercial figures are larger. Salesforce says agent users spend 8% more than the average buyer and purchase 23% more items per order. Brands using AI agents are also seeing seven times faster sales growth than brands without them.

ZDNET's summary does not say whether that seven-times comparison covers the same period, products or markets. Nor does it show whether agent users would have made the same purchases without an agent. The claimed loyalty and sales gains remain strong correlations, not verified causal effects.

The survey also distinguishes reach from value in its own framing: 39% used an agent on an e-commerce site, while referral users show higher purchase rates and order sizes. It does not disclose the base sizes for each subgroup, however, so the percentage gains cannot be translated into shopper counts or revenue without inventing precision.

Retailers would still need to service that traffic. Salesforce's report states: "Retailers must adopt AI shopping agents in order to compete." It says retailers must also service and respond to shopper agents.

The required capability is agentic engagement: detecting intent, buying signals, sources of dissatisfaction and other buying patterns at machine scale, meaning faster and at higher volume than human service.

Agent referrals may increase order value, but retailers would also carry the service burden. The survey supports an association between agent use and spending, not a causal sales lift.

Why won't shoppers let agents spend?

Adoption does not equal authority. According to Salesforce, only 19% of personal-agent users use an agent to check out. Across consumers, 73% say they are uncomfortable giving personal agents payment information, and 49% are not interested in agents making purchases on their behalf.

Consumers are not refusing every financial connection. Salesforce found that 9% allow unrestricted access to their financial details, while 36% have set up a separate credit card for their AI agents. The report says shoppers also resist sharing purchase history or personal data that could potentially lead to unwanted charges.

The 36% using separate cards represents a middle path between unrestricted access and complete refusal. Salesforce does not report typical spending limits for those cards or whether they prevent charges outside a shopper's intended budget.

Reliability has a short fuse in the survey. Salesforce says shoppers will abandon an AI agent after 2-3 mistakes. If an agent makes a bad recommendation, 1 in 3 consumers will blame the AI.

The source does not define what counts as a mistake or separate technical failures from poor product judgment. Payment permission is the immediate bottleneck, and shoppers retain the right to abandon an agent after repeated errors.

Will autonomous checkout arrive soon?

Salesforce expects AI agents to move from advising consumers to making purchasing decisions with little or no human involvement in the near future. Its report predicts: "The future of shopping will be more autonomous, highly personalized, and nearly frictionless."

That forecast sits uneasily with current behavior. Only 19% of personal-agent users use agents to check out, 73% of consumers resist sharing payment information, and 49% express no interest in delegated purchases. Those figures support today's advisory model and limited delegated transactions, not frictionless autonomy.

Salesforce gives no adoption curve, target date or forecast method for its near-term prediction. Nor does it provide implementation costs, technical standards, security controls or retailer case studies that would show how brands convert agentic AI investment into the promised loyalty.

Technical readiness may be the easier half of the problem, but the evidence remains thin. The source gives no definition of an autonomous purchase, threshold for human oversight or account of returns, disputes and unauthorized transactions. Those missing controls matter because retailers need consumer trust to gain checkout authority.

Salesforce expects brands investing in agentic AI to gain the most loyalty and future business. That remains a forecast, and the source offers no way to test its scale.

For retailers, agent readiness may determine whether AI remains a referral channel or becomes checkout infrastructure. For consumers, it decides whether convenience comes with delegated control of money. Salesforce's next claim to test is simple: whether agents gain checkout authority faster than they lose user trust.

Original: salesforce.com

Share this article:

More from James Calloway

James Calloway

Show full bio

News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.

749 articles

Related articles

  1. AI Could Soon Set the Price of Your Big Mac and Groceries
  2. Bloomreach acquires AthenaHQ to connect AI search and personalization
  3. 4.5% of US AI Users Pay — and the Top 1% Spend $900 a Month
  4. Meta Claims Millions Shop With Muse. Executives See 'Hundreds' of Orders
  5. AI Search Referrals Leave Retailers A$43,400 Short, Model Shows

« Previous articleNext article »