Who wins and loses as AI agents reshape shopping?

A Boston Consulting Group report, based on a survey of 2,500 US consumers "broadly representative of the US population," argues that agentic commerce will produce clear winners and losers across the shopping value chain: "consumers and networks standing to gain most, merchants facing fundamental challenges, and issuers presented with both risks and opportunities." BCG reports that 53% of consumers have already used GenAI tools or chatbots, 81% expect to use GenAI tools in shopping for at least one product category, and 42% say they would allow a GenAI agent to shop entirely on their behalf for at least one category. Against $2.3 trillion in US e-commerce spend in 2024, BCG estimates "over $1.3 trillion will be impacted by agentic commerce, of which almost $400B will be mostly autonomous," later refining this to 55% of e-commerce spend becoming agent-assisted and 18% autonomous, equating to the same $1.3 trillion figure.

BCG frames the pace of adoption as governed by supply and demand constraints rather than by capability alone. On the supply side, products that are complex, risky, slow to deliver, low in online availability, or luxury (the report cites cars and designer jewelry) will lag. On the demand side, the report names "a lack of customer education around AI agents and low levels of trust initially, exacerbated by news flow in relation to hallucinations and data security," plus resistance to provider monetization such as added subscriptions or fees. BCG also reports that consumers judge whether to grant an agent purchase autonomy chiefly on familiarity with the category, the ability to return or cancel for a full refund, and lower product cost — and that the categories consumers see as readiest for agentic adoption are household supplies, food delivery, hygiene and beauty products, over-the-counter medicine, and pet food.

This sits alongside other consultancy surveys already in the library. Do consumers actually want to delegate shopping decisions to AI? reports higher headline trust but similarly selective delegation; BCG's 42% willing-to-fully-delegate figure, drawn from a different survey population and framed around specific product categories rather than a general trust question, is a lower-magnitude but directionally consistent data point. Do consumers want to approve AI shopping purchases before they happen? names pre-purchase approval as the top safeguard consumers want; BCG's finding that return/refund ability and category familiarity (not approval itself) drive the autonomy decision adds a different axis — the conditions under which approval becomes unnecessary, rather than how approval should work. Where What drives shoppers to accept autonomous AI assistants? isolates perceived benefit as the adoption driver, BCG's supply/demand framing adds concrete frictions — hallucination concerns, data security, provider monetization — that could explain why perceived benefit does not always convert into adoption. The same report's own evolutionary staging of the shopping journey is set out in How will AI agents progressively take over online shopping?.

The report is a vendor survey, not peer-reviewed research, and BCG does not publish the survey instrument or define precisely how "impacted," "agent-assisted," and "mostly autonomous" spend were operationalized or modeled. The figures describe stated intent and a forecast, not observed purchasing behavior — nothing here shows agents actually executing purchases at these volumes today. The winners/losers framing is BCG's own strategic argument aimed at payments-industry readers, and its incentive to warn merchants and issuers to act now should be weighed against how firmly the figures are evidenced.