Agentic shopping is the use of AI agents to carry out parts of the shopping process on behalf of the customer instead of merely providing recommendations. Google’s agentic shopping protocol, for instance, enables agents to work across different parts of the buying processes, including discovery and post-purchase support. But has agentic shopping reached the capacity to directly make shopping decisions on customers’ behalf?
AI agents are increasingly able to independently choose products, make payments, and recommend financial options. However, a recent Bernstein note describing the outcome of a test of five different agentic shopping tools says that “until payment is embedded directly into the experience, AI remains one step removed from making shopping decisions on behalf of users.”
The firm tested ChatGPT, Gemini (NASDAQ: GOOG), Claude, Amazon’s Alexa (NASDAQ: AMZN) and Walmart’s Sparky (NASDAQ: WMT).
The goal was to see what approach the agentic shopping tool took to the shopping funnel: product identification, pricing validation, and finally purchase enablement. The analysts assessed where each tool breaks down, and how far we might be from a “true end-to-end agentic e-commerce experience.”
Bernstein says the key difference between general AI tools and retail-native AI tools built on top of retail ecosystems is emphasized by their particular performance gaps. Foundational AI models not tied to a retailer seem to provide freewheeling results from a variety of web sources, compare products across platforms, and identify deals in a fragmented retail landscape.
But they also have a distinct limitation compared to retail-native AI in that they do not have information about structured catalogue data, real-time pricing, inventory, and delivery options.
On the other hand, retail-native AI agents provide a highly precise and almost transaction-ready experience, but with results limited to their own assortment.
Bernstein says, “Bridging this gap, from either end, will be key as the technology evolves.”
The big learning of the test was that “there is currently no agentic shopping tool capable of buying a product in an unsupervised fashion. Human engagement is needed at many steps.”
ChatGPT, Claude, and Gemini don’t have access to retailers’ assortment and rely on workarounds like web-scraping and third-party articles. This means they struggle with SKU identification and pricing accuracy sometimes. The analysts say they also fail when it comes to handling fulfilment and payment.
In the final analysis, ChatGPT, Claude, and Gemini are product discovery tools, and their performance can be uneven. However, they are in the process of partnering with retailers to address these limitations.
ChatGPT’s Agentic Commerce Protocol [ACP], and Claude’s integration with Uber Eats (NYSE: UBER) are steps in this direction, making it possible to order groceries and food deliveries. However, as of now, transactions still need to be completed by customers following links to the retailers’ websites.
The analogue of that limitation is also present on Amazon’s Alexa and Walmart’s Sparky. Transactions cannot be completed directly within the assistant.
source https://www.investing.com/news/stock-market-news/how-agentic-is-ai-shopping-today-4741112
