How Walmart Is Using AI to Lift AOV, Conversion, and E-Commerce Growth

Walmart’s AI strategy is not just one tool

The headline product is Sparky, Walmart’s shopping assistant. But the more important signal is structural: Walmart seems to be building an AI stack that touches discovery, conversion, basket building, fulfillment, and external distribution through third-party agent ecosystems.

Based on the available context, that stack includes:

• A customer-facing assistant in Sparky
• Visual search to improve product discovery
• AI-driven operational improvements tied to fulfillment and execution
• Deep links and integrations with outside AI ecosystems
• An open posture toward agentic commerce, including partnerships with emerging shopping agents

This is a different approach from retailers that are trying to keep AI agents at arm’s length. Walmart appears to be betting that if shopping behavior shifts toward delegated purchasing, being accessible to those agents could become a demand advantage.

The measurable retail outcomes are the real story

Many AI commerce announcements stay vague. Walmart’s case stands out because the described impact is tied to familiar retail metrics.

The context points to three signals in particular:

• Customers using Sparky generated average order values 40% higher than other shoppers
• Visual search drove a 57% increase in conversion compared with text-based search
• Weekly engagement with Sparky rose 60% quarter over quarter

There is also a broader business outcome attached to this effort. Walmart’s e-commerce sales were described as growing 24% year over year, with management linking part of that performance to AI-driven efficiencies, faster delivery, and reinvestment into pricing and customer experience.

The careful reading here is important. AI is likely not acting as a single growth lever. It is contributing across several points in the funnel:

• Better discovery increases conversion
• Better assistance can increase basket size
• Better fulfillment improves customer experience
• Better economics create room to compete on price and speed

That combination is harder to replicate than a standalone assistant.

Why visual search matters more than it sounds

Visual search can look like a convenience feature, but in retail it solves a specific problem: many shoppers do not know the exact words for what they want. They may know the style, shape, color, use case, or a rough mental image.

Text search breaks down in those situations. Visual search shortens the path from vague intent to relevant products.

If Walmart is seeing materially higher conversion from visual search than from text-based search, the implication is straightforward. AI is not only helping customers ask better questions. It is helping Walmart reduce search friction at the moment where purchase intent is still unstable.

For retailers and tool builders, this is a useful pattern to watch. Search is no longer just about query matching. It is becoming an intent interpretation layer.

Sparky’s AOV effect suggests higher-value shopping journeys

The 40% higher average order value tied to Sparky users is especially notable because AOV is not only a monetization metric. It often reflects shopping quality.

A higher basket can come from several useful behaviors:

• Better cross-sell and complementary product discovery
• Reduced uncertainty during decision-making
• More confidence in larger or more complete purchases
• Smoother navigation through a large catalog

In other words, the assistant may be doing more than answering questions. It may be helping customers construct a fuller purchase.

That distinction matters in AI commerce. Plenty of assistants can retrieve information. Fewer can guide a shopper toward a better basket without creating obvious friction or distrust.

Walmart is also playing the distribution game

One of the more strategic details in the context is Walmart’s openness to external agents and AI ecosystems. That includes third-party assistants and shopping agents that can direct users toward Walmart when relevant context exists.

This is where Walmart’s position starts to look less like a customer service initiative and more like channel strategy.

If AI agents become a meaningful interface for commerce, retailers may face a new platform question:

• Do they block these agents to protect direct control?
• Or do they cooperate and try to become the preferred merchant inside agent-led shopping flows?

Walmart appears closer to the second camp. Amazon, by contrast, is presented as taking a more restrictive approach toward at least some external shopping agents.

Neither position is obviously wrong.

Blocking protects the merchant’s funnel, customer data, and interface control. It also reduces the risk of handing power to a new intermediary.

Opening up may create exposure to new demand flows, especially if consumers begin outsourcing product comparison, cart assembly, and checkout preparation to AI systems.

The tradeoff is clear: control versus reach.

Agentic commerce is the bigger trend behind this

The Walmart story is really a trend story about agentic commerce.

Agentic commerce shifts the center of gravity from manual browsing to delegated intent. Instead of visiting ten product pages, a customer may ask an agent to compare options, narrow the field, prepare an order, and route the purchase to the best merchant that meets price, speed, trust, or convenience requirements.

In that world, a retailer has to optimize for two audiences:

• Human shoppers
• Machine-mediated shoppers

That changes what “good e-commerce” looks like.

A strong retail AI strategy may increasingly require:

• Structured product data
• Reliable availability and fulfillment signals
• Smooth deep linking or API access
• Clear checkout handoff paths
• Search and recommendation systems that interpret fuzzy intent

Walmart seems positioned around this logic. Its internal assistant improves the direct experience, while its openness to partnerships gives it a foothold in external shopping agents-led demand.

The ad model question is still unresolved

There is one major uncertainty in agentic commerce that should not be ignored: advertising.

If more shopping decisions happen through screenless or semi-automated interfaces, the traditional retail media model becomes less stable. Sponsored placements, display formats, and attention-based merchandising all assume a visible interface where people browse, compare, and click.

If an agent is making or narrowing decisions upstream, several questions follow:

• What becomes the equivalent of shelf placement?
• How are sponsored results disclosed and ranked?
• Does the retailer, the agent platform, or both capture the advertising economics?
• How much influence will brands retain over discovery?

This matters for Walmart because advertising is increasingly important to major retailers. A more agent-driven shopping environment could help conversion while simultaneously pressuring existing retail media assumptions.

That is not a contradiction. It is a real strategic tension.

Why Walmart’s approach looks pragmatic

What stands out is not blind optimism about AI agents. It is optionality.

Walmart appears to be improving the direct shopping experience today through tools like Sparky and visual search. At the same time, it is not closing itself off from a future in which external agents become influential demand brokers.

That is a practical stance because no one yet knows how far consumer shopping agents will go. They may become mainstream for repetitive purchases, high-friction comparison tasks, or bundled planning scenarios. They may remain niche in more emotional or discovery-driven categories.

By building for both possibilities, Walmart reduces the risk of choosing the wrong interface future too early.

What founders, retailers, and AI tool buyers should take from this

For anyone tracking AI commerce, Walmart’s example highlights a few useful lessons.

The practical takeaway

Walmart’s AI story is less about one assistant and more about a retail architecture designed for both current shoppers and emerging shopping agents. The immediate gains appear to be in conversion, average order value, and e-commerce momentum. The longer-term bet is that openness to agentic commerce could turn into a distribution advantage.

For teams evaluating AI in retail, that is the key lens to use: not “Do we have an AI feature?” but “Are we making it easier to buy, easier to fulfill, and easier to be chosen by the next layer of commerce interfaces?”