What Walmart's Sparky AI Earnings Results Prove About Conversational Commerce and AOV Lift

Walmart disclosed something specific on their earnings call that most operators glossed over.

Shoppers who used Sparky (Walmart's in-house shopping AI agent) placed orders with an average order value about 35% higher than shoppers who didn't. That figure came from Walmart's own first-party data, reported by Digital Commerce 360 in May 2026. It's not a projection or a pilot result. It's a disclosed outcome from one of the highest-volume retailers on the planet.

That number deserves more attention than it's gotten.

What Sparky Actually Does

Sparky is not a search bar with a chat interface bolted on. It's an action-capable agent. It understands complex, multi-part queries, reasons across Walmart's product catalog, makes comparisons, and helps a shopper move toward a decision, not just a results page.

That distinction matters. A static search returns a list. Sparky holds a conversation. It handles the kind of question a shopper actually has: "I need a patio set that seats six, fits under a 10-foot pergola, and ships before the long weekend." A keyword search returns every patio set. An action-capable agent narrows, clarifies, and guides.

The AOV lift follows directly from that capability. When a shopper can ask a complex question and get a useful answer, they buy more confidently. They add the matching cushions. They take the extended warranty. They don't abandon the session because the results page confused them.

Why Conversational AI Lifts AOV

Two mechanisms are at work, and they compound.

Friction reduction. Static product pages are built for the median shopper. They answer the questions the category team anticipated, not the question this specific shopper has right now. When a shopper hits that wall, they either leave or they settle for a less confident purchase. Conversational AI removes the wall. The shopper gets an answer, keeps moving, and buys with more certainty.

Scope expansion. A conversation is a two-way surface. A static page can only show what it was built to show. A conversational agent can ask a clarifying question and surface a product the shopper didn't know to search for. That's not upselling in the manipulative sense. It's accurate matching. When the match is better, the basket is larger.

Walmart's 35% AOV lift reflects both mechanisms operating at scale. Shoppers who used Sparky weren't a self-selected group of high-spenders. They were shoppers who engaged with a conversational interface and, as a result, bought more.

The Gap Between Walmart and Everyone Else

Here's the problem. Walmart built Sparky over years, with a team and infrastructure that no mid-market DTC operator has access to. Building an action-capable shopping agent from scratch means integrating your product catalog, your inventory system, your checkout flow, and a large language model that can reason across all of it, then keeping it current as your catalog changes.

That's not a six-week project. For most operators, it's not a realistic project at any timeline.

So the Walmart result sits in a frustrating position: it proves the principle works at scale, but the path to replicating it looks closed to anyone without Walmart's resources.

That framing is worth questioning.

The Principle Is Portable. The Build Is Not.

The underlying mechanism (a conversational layer that handles complex queries, reduces friction, and guides shoppers toward confident decisions) doesn't require a Walmart-scale engineering team. It requires the right architecture.

This is where the question shifts from "can we build a Sparky?" to "can we deploy the same capability without building it from scratch?"

The answer depends on what you're willing to accept as a starting point. If you need a fully custom agent trained on proprietary signals and integrated into a legacy commerce backend, you're back to the multi-year build. But if your goal is a conversational, action-capable surface that handles the queries your static pages can't and routes shoppers toward checkout. The barrier is much lower.

Aigency builds exactly this kind of surface. It creates a parallel, agentic version of your site on a brand-owned subdomain (ai.yourbrand.com), using vision-based agents that crawl your existing site without requiring any code changes. The result is a site that can hold a natural language conversation with shoppers, answer product questions, and guide them toward a purchase decision.

The vision-crawl approach matters here. Aigency reads your site the way a human would, by looking at it, rather than requiring API access to your commerce backend or a custom data feed. You don't have to replatform. You don't have to rebuild. The agentic layer sits alongside your existing site and handles the traffic your static pages were never designed to serve.

The Traffic Pattern That Makes This Urgent

The Walmart result is notable on its own. It becomes more urgent when you layer in what's happening to traffic sources.

Adobe Analytics reported that AI-driven referrals to retail sites grew roughly 1,300% year over year through early 2026. Shoppers are arriving at your site after a conversation with ChatGPT, Perplexity, or Gemini. They've already asked a complex question. They've already received a partial answer. They're arriving mid-conversation, not at the start of a browse session.

Your static product page was built for a shopper who starts from zero. It wasn't built for a shopper who arrives asking: "The AI said you carry a version of this that works with my existing system. Does it?"

That question lands on a page that can't answer it. The click happened. The sale did not.

This is the same gap Sparky closes for Walmart. The shopper has a specific, contextual question. The static page returns a generic answer. The conversational agent handles the question and keeps the session alive.

What the Sparky Result Actually Tells You

The Walmart disclosure is useful not because you should try to build Sparky, but because it confirms the mechanism with real data at real scale.

Conversational, action-capable AI lifts AOV. It does so because it reduces friction on complex queries and expands the scope of what a shopper considers. Those aren't Walmart-specific dynamics. They apply to any shopper who arrives with a question your static pages can't answer.

The question for your operation isn't whether the mechanism works. Walmart's earnings call settled that. The question is whether you have a surface that can apply it.

Aigency's free site scan surfaces the specific questions AI-referred visitors are already asking that your current pages fail to answer. That's a useful place to start, regardless of what you decide to do next.

See what your site is missing at aigency.ai.

FAQs

What did Walmart's Sparky AI results actually show about AOV? According to Digital Commerce 360's May 2026 report on Walmart's earnings disclosure, shoppers who used Sparky placed orders with an average order value about 35% higher than shoppers who did not. This figure comes from Walmart's own first-party data.

Why does conversational AI increase average order value? Two mechanisms drive AOV lift. First, conversational AI reduces friction by answering the specific question a shopper has, rather than returning a generic results page. Second, it can surface relevant products the shopper didn't know to search for by asking clarifying questions. Both effects lead to larger, more confident purchases.

Can a mid-market DTC brand replicate what Walmart built with Sparky? Building a fully custom agent like Sparky from scratch isn't realistic for most operators. It requires deep integration across catalog, inventory, and checkout systems, plus sustained engineering investment. The underlying capability (a conversational layer that handles complex queries and guides shoppers to checkout) can be deployed without a custom build using purpose-built agentic storefront infrastructure.

What is an agentic website and how is it different from a chatbot? An agentic website is a full, conversational version of your site that can understand complex queries, reason across your product catalog, and guide shoppers toward a purchase decision. A chatbot typically handles a narrow set of scripted questions. An agentic layer handles open-ended, contextual conversations and connects directly to your actual product and checkout flow.

Who are the AI-referred visitors that need a conversational experience? These are shoppers who arrive at your site after interacting with an AI assistant like ChatGPT, Perplexity, or Gemini. They've already asked a detailed question and received a partial answer. They arrive mid-conversation, expecting to continue that conversation on your site. A static product page wasn't built for that entry point.

Does deploying an agentic layer require changes to my existing site? Aigency uses vision-based agents that crawl your existing site without requiring code integration or changes to your commerce backend. The agentic layer lives on a brand-owned subdomain alongside your current site.

How do I know if my site has a gap for AI-referred visitors? Aigency offers a free site scan that surfaces the specific questions AI-referred visitors are already asking that your current pages fail to answer. Run it at aigency.ai to see where the gaps are before deciding on next steps.