When AI does our shopping for us

Berlin, October 8, 2026
 

Agentic Commerce – the use of AI agents in online shopping – is expected to develop gradually from the consumers’ perspective. A majority are open to the use of AI in shopping and are already using such applications today. However, the willingness to hand over the entire shopping process to an AI agent remains limited. Support with research, product comparisons and the evaluation of alternatives is particularly well accepted. This is shown by the latest KPMG Consumer Snapshot 03/2026, which was conducted and analysed by the EHI Retail Institute on behalf of KPMG.

46 per cent have used AI frequently or very frequently in purchasing decisions over the past twelve months, whilst 73 per cent of respondents can envisage using an AI agent to compare products, prices or offers. There are differences between respondents in this regard: men use AI slightly more frequently than women when making purchasing decisions, and its use is significantly more widespread in urban areas than in rural ones.

The specific shopping situation also influences what consumers would use an AI agent for. This is particularly evident in the case of smartphones: although choosing a product involves a high financial risk and a considerable amount of research, 72 per cent would seek assistance from an AI agent when searching for information. When it comes to a binding step such as payment, however, this willingness drops significantly to 45 per cent. The results suggest that consumers today primarily wish to use AI agents for research and guidance, whilst they tend to want to retain control themselves over final steps that are difficult to rectify.

From the consumers’ perspective, delegating shopping activities to AI agents raises significant concerns. The greatest risks are considered to be the systematic favouring of certain retailers and the unauthorised use of data from search criteria, personal preferences and the payment process. Added to this are unreliable information, unsuitable product recommendations and the potential loss of control over the purchasing process. Retailer and manufacturer AI faces a particular challenge in building trust: 64 per cent prefer general generative AI when it comes to objectivity and independence from commercial interests, whilst only 7 per cent prefer retailer or manufacturer AI.

79 per cent want to be able to intervene and make decisions themselves at any time, whilst 74 per cent want to determine for themselves which personal data the AI agent uses. For 72 per cent, the ability to set rules, budgets and exclusion criteria is a key requirement. Furthermore, 67 per cent want transparency regarding whether suppliers’ interests are taken into account in recommendations.

Agentic Commerce can establish itself in the long term if AI agents offer recognisable benefits, provide clear justifications for recommendations, and ensure their actions remain transparent, verifiable and correctable. Further development will therefore be determined less by the maximum possible autonomy of the systems than by clear rules governing the sharing of decision-making.

When an AI agent is involved in selecting, preparing or processing a purchase decision, consumers prioritise product quality, the lowest price, and the reliability and trustworthiness of the retailer and the product. A quick, clear overview of the most important product information also ranks highly, whilst customer reviews, short delivery times and the speed of the purchasing process are considered less important. Sustainability, consideration of a general brand preference and the speed of the purchasing process are among the lowest-priority criteria.  In a direct comparison of the criteria surveyed, sustainability and brand preference are less frequently cited as key factors.

The new focus study is available to download free of charge. The digital version of the Consumer Snapshot is available via a free subscription: www.kpmg.de/consumerbarometer.