Walmart operates the largest retail catalog in the world. Hundreds of millions of products, each with descriptions, attributes, images, categories, and search metadata — data that determines whether a customer finds what they’re looking for, whether inventory gets to the right warehouse, whether a supplier gets the right order.
In August 2024, CEO Doug McMillon made a statement during the Walmart Q2 FY2025 earnings call that reframed how to think about AI in large-scale operations. Walmart had used generative AI to improve over 850 million data points in its product catalog — work that would have required 100 times the human workforce to complete manually in the same timeframe.
The original insight: some tasks are only possible with AI
The standard framing for AI ROI is efficiency: AI makes a process faster or cheaper. Walmart’s catalog decision was different. The scale — 850 million data points — was not a task where AI accelerated an existing process. At that volume, manual completion in any reasonable timeframe was operationally out of reach. AI did not make the task faster. It made the task feasible at all.
McMillon’s framing was direct: “the quality of the data in our catalog affects nearly everything we do, from helping customers find and buy what they’re looking for, to how we store inventory in the network, to delivering orders.” The AI transformation was not a productivity initiative. It was a data quality initiative — at a scale that only generative AI could execute [Retail Dive, August 2024, citing Q2 FY2025 earnings call].
The results — verified from earnings call disclosures
- 850 million product catalog data points improved with generative AI — CEO Doug McMillon, Q2 FY2025 earnings call [Retail Dive]
- Productivity multiplier: 100x compared to manual process for the same output [Modern Retail]
- 30 million unnecessary delivery miles eliminated with AI route optimization
- 94 million pounds of CO₂ avoided through optimized delivery routing
- Walmart offered the route optimization technology as a SaaS solution to all businesses as of March 2024
What changed operationally
Improved catalog data quality cascades through every layer of Walmart’s operation. Associates can locate inventory faster via mobile tools. Customers receive more accurate search results. Delivery routing becomes more precise because the underlying product data is cleaner.
Before the AI improvement, incomplete or inaccurate product data created friction at every touchpoint downstream: a customer search that returns the wrong product, a warehouse that stores inventory in the wrong location, an order that routes to the wrong fulfillment center. The AI did not add a new capability — it removed a systemic defect at scale.
The route optimization result — 30 million delivery miles eliminated, 42,000 tons of CO₂ avoided — came from a parallel AI initiative using machine learning to optimize delivery paths across Walmart’s supply network. The performance was strong enough that Walmart packaged the technology as a commercial SaaS offering.
What you can take from this
The 850 million data points story offers a useful reframe for how to identify AI opportunities: look for tasks where the scale is the constraint, not the complexity.
Many AI pilot programs target complex tasks — judgment calls, creative work, nuanced decisions. Those are harder to automate. The category where AI delivers the clearest enterprise value is often the opposite: high-volume, structured, repetitive tasks where manual effort cannot scale to match the problem size.
Walmart’s catalog was not too complex to do manually. It was too large. AI did not need to be smarter than a human cataloguer — it needed to work at 100x the volume. That is a very different problem to solve, and a much easier case to build.
SAGA — Success Stories & Real Cases · Curating real AI implementations: the original idea, the decision, the verified result.