Building an AI Operating System for Retail, with Walmart VP Anupriya Sharma

Walmart describes its technology stack as something closer to an operating system, one designed to connect customers, merchants, stores, and supply chains through a single AI layer.

For Anupriya Sharma, Vice President of U.S. Technology Strategy, Portfolio, and Technical Program Management at Walmart, the shift is part of a change in how the company views the role of software inside the business.

“Walmart’s evolution from a data-driven retailer to an enterprise powered by agentic AI reflects a major shift: from reactive tools to proactive, task-orchestrating agents,” Sharma said. “While our people-led, tech-powered philosophy endures, the biggest shift is the move from reactive digital tools to proactive agents like Sparky and our Merchandising Agent.”

In a June 2025 product announcement, the company described Sparky (its consumer-facing AI assistant) as a foundation for more proactive, conversational shopping experiences inside the Walmart app, including product discovery, recommendations, and planning workflows such as recipe-to-cart journeys.

Sharma framed Sparky and Walmart’s internal tools as complementary expressions of the same design principle, saying that they exemplified Walmart’s core belief: combining AI precision with human judgment to remove friction, simplify experiences, and empower people.

AI on Both Sides of the Transaction

Sparky is part of a shift toward conversational interfaces in retail, an area where Walmart has accelerated development over the past two years. The assistant lives inside the Walmart app and is designed to synthesize reviews, answer natural-language queries, and guide customers through product selection. Walmart has said the system is built on proprietary retail data layered with large language models.

On the enterprise side, Walmart rolled out an internal generative AI tool for merchants in early 2025, often referred to publicly as “Wally,” to help teams analyze assortments, performance data, and operational questions without relying on traditional dashboards or manual reporting. Sharma referred to this class of tooling more broadly as a Merchandising Agent, emphasizing its role as an access layer rather than a single feature.

“Products like Sparky and our Merchandising Agent, built on our proprietary data, automate work, deliver deeper insights, and are now widely used by merchants, including on the mobile,” Sharma said. “We invest where AI improves decision-making and precision, from merchandising to our real-time supply chain.”

That focus on compounding impact, rather than novelty, shapes how Walmart evaluates emerging technologies at scale. With people across the U.S., India, and Israel working within its technology organization, the challenge is filtering signal from noise.

“At Walmart’s scale, the challenge isn’t chasing every signal, it’s identifying the ones that will compound,” she said. “We saw Generative AI not just as a better chatbot, but as the foundation for agentic AI, enabling a shift from searching to contextual shopping.”

Public disclosures support that framing. Walmart’s most recent annual filings highlight continued investment in automation, AI, and data platforms as part of its long-term strategy, rather than isolated pilots. The company has also continued to expand Global Tech hiring tied to AI infrastructure and applied machine learning.

How Walmart Uses AI to Reduce Uncertainty

The clearest operational impact of Walmart’s AI systems may be in logistics and fulfillment, where the company has spent years building software to coordinate store-based delivery at national scale. Walmart now fulfills a majority of U.S. e-commerce orders from stores, a model that depends on accurate inventory visibility and delivery timing.

“At Walmart, AI is embedded throughout the enterprise versus applied in isolation,” Sharma said. “Our focus is on using AI to turn uncertainty into precision across the customer experience.”

Delivery windows offer a clear view into how AI is now embedded in Walmart’s operations. What were once static estimates have been replaced by systems that update in real time as conditions change across fulfillment, transportation, and stores.

These models continuously adjust delivery commitments using large volumes of operational signals, with the aim of increasing reliability rather than simply forecasting outcomes. The emphasis is on execution and customer confidence, ensuring orders arrive when promised as the system learns and recalibrates.

Sharma also drew a distinction between what generative AI adds to the shopping experience versus earlier generations of machine learning.

“By anticipating customer needs and understanding intent, GenAI creates a more seamless connection between customers and Walmart’s assortment.”

That distinction mirrors industry trends, as retailers experiment with conversational commerce and AI-assisted browsing. Walmart was among the first major retailers to test direct integrations with OpenAI-powered shopping experiences in 2025, signaling a willingness to expose parts of its catalog to third-party AI systems while keeping core decision-making in-house.

Building the Retail Operating System

Walmart’s technology organization has invested in a digital backbone that links data, automation, and AI capabilities across stores, fulfillment centers, and apps, creating an architecture that supports connected decision layers rather than isolated systems. Independent analysis of the company’s AI strategy also notes efforts to consolidate AI tools into unified frameworks that span customer, associate, supplier, and engineering domains.

“Building a unified, intelligent retail operating system goes beyond any single technology. It’s about integration, trust, scale, and governance,” Sharma said. “It starts with high-quality, governed data that can be responsibly shared across systems operating at different speeds and complexities.”

Governance has become a more visible part of Walmart’s AI messaging as systems move closer to customer commitments and associate workflows. The company maintains public principles around responsible AI use and model oversight, emphasizing explainability and resilience in high-impact systems.

“Strong AI governance is essential to ensure models are ethical, transparent, explainable, and resilient, especially when they impact customer commitments and associate workflows,” she said. “Walmart is uniquely positioned to create an AI-powered operating system that is innovative yet dependable. Scalable enough to serve millions daily while earning trust through responsible data use and reliable execution.”