AI Agents and Ecommerce: The Distrust (and ROI) Numbers Salesforce Just Published Ahead of the 2026 Holidays

On November 12, 2026, Salesforce published on its official site (salesforce.com/news) the results of a Consumer Shopping survey of more than 2,500 shoppers in the United States, United Kingdom, Canada, and Australia. The company did not disclose fieldwork dates or margin of error. The takeaway cuts both ways: shoppers remain largely distrustful of letting AI agents pay on their behalf, but brands that deploy their own shopper agent report better business results.

Distrust Still Dominates on Payment

• 73% of surveyed shoppers say they're uncomfortable giving a personal AI agent their payment information.
• 49% have no interest in letting an agent browse or buy on their behalf.
• Only 9% of personal agent users give it unrestricted access to their financial details; 36% prefer to dedicate a separate card to it.
• 31% of shoppers have tried a personal AI agent — Meta Muse being the most used (12%) — and among them, only 19% have used it at checkout so far.

When the Agent Gets It Wrong, Patience Runs Out Fast

• Per Salesforce, most shoppers abandon an AI assistant after just two or three mistakes.
• 32% of shoppers blame the AI when it makes a bad recommendation.
• Among personal agent users: 27% report the agent triggered a purchase that was too expensive, 22% had the agent blocked before completing a purchase, and 32% report no major errors.

The Counterpoint: Brands With a Shopper Agent Sell More

• 39% of shoppers have used a brand's shopper agent (as opposed to a personal one); 80% describe a good experience, and 75% of those say they feel more loyal to the brand since.
• Shoppers who engage with a brand's shopper agent spend an average 8% more per order and buy 23% more items, per Salesforce.
• Brands that have deployed an AI shopping agent see their sales grow, per the vendor, 7 times faster than those that haven't — with no detail on sample size or calculation method.
• Ecommerce traffic from AI tools is up 104% year over year, and AI-referred shoppers convert more than twice as often as typical ecommerce shoppers.

The Limits Worth Knowing Before Drawing Conclusions

• The survey discloses no fieldwork dates, no margin of error, and no country-by-country breakdown — making the statistical robustness of the figures hard to assess.
• Results are self-reported by surveyed shoppers, with no independent measurement.
• The survey's publisher (Salesforce) also sells shopper agents and Agentforce for commerce — worth keeping in mind when reading the most favorable figures, especially the "7x faster" growth claim.
• The $75.8 billion projection for AI-driven Cyber Week spending (+13% vs. 2025) is a Salesforce proprietary estimate, not an independent market measurement.

What This Means for Your Ecommerce Site Ahead of the Holidays

• Prioritize product data: accurate prices, stock, and product pages are the precondition for any AI agent (personal or third-party) to correctly recommend your offer — Salesforce explicitly says so in its release.
• An in-house shopper agent carries more perceived value than a generic third-party agent in the survey (39% usage, 80% good experiences): if it's on your roadmap, scope it to handle complex questions rather than basic FAQs.
• Plan for additional non-human traffic (personal agents, AI crawlers) on your infrastructure ahead of the year-end peak, without degrading the experience for human shoppers.
• With 73% of shoppers still reluctant to pay through an agent, agentic payment remains a medium-term bet for most SMBs/mid-market companies — the immediate priority is product data quality, not integrating an agent payment protocol.

Wondering whether your product data and Salesforce setup are ready for the AI-driven shopping season? Our Salesforce consulting offering audits your product data and current configuration before you consider a shopper agent.