Walmart
genAI shopping assistant
Walmart customers who use the Sparky genAI assistant have an average basket about 35% higher than non-users, and units purchased through Sparky more than quadrupled in one fiscal quarter (FY2026 earnings call).
Key points
- Sparky genAI shopping assistant, from intent-based search to recommendations.
- In-house LLM and RAG setup over the catalog and customer reviews.
- Average basket +35% vs non-users, units purchased more than quadrupled.
- Evidence level A, confirmed status, present on the US app, site, and stores.
Objective
Move from keyword search to intent-based search to increase average basket and app usage frequency.
The deployment
Sparky is Walmart's genAI assistant, accessible via the Ask Sparky button in the app and then extended to the site and stores. The customer expresses an intent (organizing a birthday, choosing a sunscreen under 30 dollars) and Sparky returns recommendations synthesized from the catalog and reviews, with product comparisons. Launched the week of June 6, 2025 on the app. Features added later: automatic replenishment of essentials, meal planning, and Spanish support.
Results Proof A
Figures disclosed during the Walmart earnings call (fiscal year 2026), reported by Digital Commerce 360 and Modern Retail. Metrics tied to the group's financial communication.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
The stack in detail
- outil Sparky GenAI shopping assistant developed in-house by Walmart, accessible via the Ask Sparky button in the app and then on the site and in store.
- llm LLM de Sparky (modele exact non publie) Walmart does not publicly name the foundation model; the sources describe an in-house LLM + RAG multimodal setup.
- infra RAG sur catalogue et avis clients Walmart The synthesis of recommendations relies on product pages and reviews; the richness of the catalog is the limiting factor.
- plateforme App et site Walmart Exposure surfaces for Sparky, integrated into the purchase journey (comparison, add to cart, replenishment).
How it runs, concretely
For ops teams-
1Intent expression customer
The customer describes a need or an occasion via Ask Sparky.
-
2Synthesis and recommendation AI
The RAG system synthesizes reviews and product pages, returns recommendations and comparisons.
-
3Transactional actions AI
Sparky helps compare, add to cart, and, in some cases, replenish essentials.
-
4Metrics reading data team
Average basket, units, and weekly actives are tracked to steer the extension of capabilities.
The match between the intent expressed in natural language and the structured product data. If the catalog is not enriched, Sparky does not find the right product and the basket effect disappears.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- enriched and structured product catalog
- customer reviews
- intent signals per category
Org prerequisites
- in-house AI team or partner
- catalog quality governance
Possible stack
- foundation LLM
- RAG engine over the catalog
- multimodal text and image orchestration
The plan, step by step
- Step 1Audit the richness of the catalog (attributes, specs, reviews) category by category.Deliverable: Coverage report + product page enrichment plan.
- Step 2Build the catalog + reviews RAG and a first assistant on a few well-covered categories.Deliverable: Prototype that answers typical intents (occasion, budget, constraint).
- Step 3Launch the in-app beta with a dedicated entry point on a user segment.Deliverable: Average basket and engagement measures, users vs non-users.
- Step 4Add the transactional actions (comparison, add to cart).Deliverable: Transactional assistant integrated into checkout.
- Step 5Extend the capabilities (replenishment, languages) based on the metrics.Deliverable: Roadmap prioritized by average basket, units, and weekly actives.
First step: Audit the richness of the catalog (specs, attributes, reviews) before connecting an assistant, because it is the limiting factor.
The same pattern, proven in other industries
Booking.com
genAI trip planning assistant
Burberry
genAI copilot for consultants (search, image recognition, complete-the-look)
Homes.com
natural-language voice and text search, then conversational assistant
Sources
- S1 Walmart credits Sparky AI agent with lifting AOV, unit sales growth Secondary archive pending
- S2 Walmart says AI users build 35% bigger baskets than others Secondary archive pending
- S3 Walmart: The Future of Shopping Is Agentic. Meet Sparky. Primary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.