How Walmart Uses AI to Cut Costs and Boost Profit Margins

Walmart has reportedly begun limiting employees' use of an internal AI assistant called Code Puppy after demands placed on the large language model (LLM) backing the tool were significantly higher than anticipated. Initially, employees were encouraged to use Code Puppy without any restrictions or usage caps — but Walmart has now moved to assigning each employee a fixed number of AI tokens, effectively capping how much the tool can be used on a per-person basis.

Code Puppy was promoted internally as a productivity tool capable of assisting with tasks such as spreadsheet analysis, creating presentations, and other automatable workplace activities.

📋 Key Context: Walmart employs roughly 2.1 million people worldwide. Even modest per-employee AI queries can accumulate into substantial operational costs — especially as LLM providers shift from flat-rate subscriptions to pay-per-use billing models.

The change in internal policy is primarily a cost control measure. As AI inference providers increasingly move away from fixed-price, near-limitless subscription plans toward consumption-based pricing, enterprises like Walmart are feeling the financial pressure in real time.

Walmart's guidance to employees now focuses on using AI where it genuinely creates value. The company has issued direction on how workers should select the right AI tool for a given task. Employees also retain access to other AI platforms funded by the company.

💰 The Real Cost of "Token Maxxing"

Walmart had previously expanded AI tool adoption across the company and provided training for employees on how to use AI effectively — actively encouraging workers to experiment and identify successful use cases. Now that each interaction carries a direct financial cost, Walmart joins a growing list of large enterprises struggling to balance reported productivity gains against the expense of achieving them.

Part of the issue may lie in how productivity has been measured in AI-driven workflows. Tracking the number and complexity of AI tool interactions as a proxy for productivity has led many employees to game their KPIs — a phenomenon known as "token maxxing".

💬 As recently as April 2025, a partner at Sequoia Capital told The Wall Street Journal: "We all should be tokenmaxxing" — an approach that gave rise to internal AI leaderboards at companies celebrating those making the heaviest use of AI software.

Such performative practices now carry measurable financial consequences. The cost scales directly with the number and complexity of AI tasks and the specific model used. Larger thinking models — those that perform recursive reasoning steps — consume significantly more tokens per query, resulting in higher bills for end-users.

Walmart's guidance encouraging workers to choose models carefully is a direct attempt to limit spending on expensive frontier AI models for relatively low-complexity tasks, such as spreadsheet formatting and slide deck creation.

🤖 Multi-Agent AI and Hidden Cost Risks

Multi-agentic AI workflows introduce another layer of unpredictable cost. When employees initiate iterative loops running across multiple AI agents to reach a desired outcome, the cost of sub-optimal results — and the repeated refining and re-submission of prompts — becomes directly quantifiable in hard cash.

⚠️ Industry-Wide Shift: Both Anthropic and OpenAI have already migrated their higher-tier enterprise plans to per-token billing. Microsoft began charging for its GitHub Copilot development tools as of June 1st — in line with what is rapidly becoming the new financial norm across the AI industry.

The consequences are already visible at major enterprises. Uber recently disclosed that it had exhausted its entire 2026 AI budget within the first four months of the year — a stark illustration of how rapidly evolving pricing models are impacting enterprise end-users.

🎯 What Walmart's Token Cap Really Means

By implementing per-employee token limits, Walmart is pursuing three interconnected objectives:

• ✅ Controlling ongoing operational costs associated with AI inference at scale
• 🧠 Encouraging more deliberate and considered use of AI tools among its workforce
• 📊 Establishing clearer ROI metrics to evaluate the true return on its AI investments

As the AI industry matures and consumption-based pricing becomes standard, Walmart's approach may serve as a benchmark for how large enterprises recalibrate AI access policies — balancing innovation with financial accountability.

📸 Image source: Pixabay, under licence.

💡 Want to learn more about AI and big data from industry leaders?

Check out the AI & Big Data Expo — taking place in Amsterdam, California, and London. The event is part of TechEx and co-located with other leading enterprise technology conferences. Click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.