Walmart said in late July that it needed to rein in its sprawl of AI agents. After building dozens of separate tools, often accessed through different systems, the retailer admitted that the experience was getting confusing. Suresh Kumar, Walmart’s chief technology officer, described the new plan as consolidating everything into four “super agents”: one for customers, one for associates, one for sellers and suppliers, and one for developers. The logic was to stop requiring users to hop between narrow agents and instead offer single entry points that sit on top of the many task-specific agents running behind the scenes.
Just over a month later, Walmart has not slowed its pace. It has in fact introduced more agents, but in a more structured form. At its Converge 2025 technology event in Bangalore and through subsequent announcements, the company unveiled WIBEY, a developer-facing super agent, and reinforced that it already has customer, associate, and supplier super agents underway. CEO Doug McMillon told investors on the second-quarter earnings call that these agents will bring scheduling, sales data, supplier onboarding, and customer personalization into unified platforms.
Counting from July’s simplification plan to now, Walmart has committed to at least five super agents: Sparky for customers, Marty for suppliers, an Associate Agent for employees, WIBEY for developers, and another developer agent referenced by McMillon. Beneath these sit roughly 200 task-specific agents distributed across Walmart’s systems. The company’s Element machine learning platform has been upgraded to orchestrate these agents, offering stateful memory, reasoning capabilities, and cross-agent communication protocols. Walmart now treats agents the way it once treated models: catalogued, monitored, and reused across projects.
WIBEY, described internally as a “super agent for developers,” it acts as both a tool and a platform. Engineers can use it to discover existing agents, generate starter kits for new projects, and coordinate execution across Walmart’s tangle of development tools. Instead of searching portals for APIs or templates, developers can prompt WIBEY and receive a project framework tailored to their context. Sravana Karnati, Walmart’s executive vice president for global technology platforms, framed it as an “invocation layer” rather than just another dashboard. By integrating large language models, transformer systems, and the model context protocol, WIBEY lets teams automate repetitive tasks, normalize product data across geographies, and safely experiment before deploying solutions.
With this concentration of functionality in WIBEY, the company is tackling the sprawl problem head-on. Instead of dozens of scattered bots, developers have a single platform that manages context, execution, and discovery. This consolidation should make it easier to govern AI use internally, enforce compliance guardrails, and reuse solutions across teams. It also places Walmart’s developers in a position to build further agents quickly without creating yet another silo.
On the other hand, the scale is staggering. With about 200 agents active behind the scenes, Walmart is not reducing the number of agents so much as reorganizing them under fewer interfaces. Consolidation does not automatically mean simplification for governance. Each agent still carries potential risks, from errors in code automation to missteps in product categorization, and the more agents are layered together, the more complex the monitoring becomes. Walmart has added observability tools to track decision paths and reasoning steps, but whether these will be sufficient across hundreds of semi-autonomous processes is an open question.
Walmart serves more than 250 million customers weekly across 10,700 stores and 19 websites. AI agents promise to streamline customer experience, cut costs in operations, and accelerate development cycles. Sparky, the customer-facing agent, is being trained to build shopping carts automatically and suggest substitutions. Marty is being designed to help suppliers manage onboarding and campaigns. The Associate Agent pulls together scheduling, payroll, and sales data into one interface for employees. And WIBEY accelerates the work of internal teams. In each case, Walmart argues that consolidating functions into super agents is more efficient than proliferating narrow tools.
The July report framed simplification as the goal. The August launches suggest simplification means centralization of access points rather than a slowdown in new agent development. The approach could help Walmart avoid overwhelming its employees and customers with fragmented tools, but it also risks entrenching a complex web of systems that must be closely managed.
For the retail industry, the implications are significant. Walmart is setting a template where AI agents are core infrastructure. If it can keep the system coherent and secure, WIBEY and its sibling super agents could become the backbone of how a company of Walmart’s scale designs products, manages operations, and interacts with shoppers and suppliers.
