Walmart’s Sparky Delivers Transaction-Ready Precision but Lacks Assortment Breadth
WMT•A recent Bernstein test evaluated five AI shopping tools—including Walmart’s Sparky—across product identification, pricing validation and purchase enablement, finding none could complete unsupervised transactions. Retail-native agents like Sparky delivered precise, transaction-ready results limited to in-assortment items, while foundational models struggled without catalog and real-time pricing data.
1. Test of Five Agentic Shopping Tools
Bernstein evaluated ChatGPT, Gemini, Claude, Amazon Alexa and Walmart’s Sparky on their ability to handle product selection, pricing confirmation and payment integration within a simulated shopping funnel.
2. No Unsupervised Purchasing Capability
All five tools failed to complete an end-to-end transaction without human intervention, as none had embedded payment flows fully integrated into the shopping experience.
3. Retail-Native vs General AI Contrast
Retail-native agents like Sparky displayed precise SKU identification and up-to-date pricing but only within their own product assortments, whereas general AI models offered broader product comparisons yet lacked real-time catalog and inventory data.
4. Path to True Agentic E-Commerce
Bridging structured retailer data with seamless payment embedding remains the critical next step for developing AI agents capable of fully autonomous e-commerce transactions.
