There's a difference between adopting AI because it helps and adopting AI because the CEO said so. According to a Guardian investigation, Amazon is firmly in the second camp for a growing number of its internal workflows.
The report details how Amazon has pushed AI-powered tools across its operations, from warehouse logistics to corporate workflows, with mandates that employees use them regardless of whether the tools actually improve output. In several cases described in the piece, workers found the AI tools actively slowed them down compared to their previous methods.
The Mandate Problem
This isn't unique to Amazon. Across corporate America, there's a growing pattern: executives announce AI transformation initiatives, set adoption targets, and measure success by usage metrics rather than productivity outcomes. The result is employees spending time wrestling with AI tools that weren't built for their specific workflow, just so the company can report high AI adoption numbers.
The difference with Amazon is scale. When a company with 1.5 million employees mandates a tool change, the productivity cost of getting it wrong multiplies fast. A tool that wastes 10 minutes per worker per day across even a fraction of that workforce adds up to millions of lost hours.
What Actually Works vs. What Gets Deployed
The pattern across industries is consistent: AI tools work well for specific, well-defined tasks (summarizing documents, drafting initial code, sorting support tickets) and poorly for complex judgment calls or workflows that were already optimized by experienced workers.
The companies seeing real productivity gains tend to let teams choose their own tools and measure actual output changes. The ones generating headlines about "AI transformation" tend to mandate tools top-down and measure adoption rates.
Amazon has the engineering talent and resources to build genuinely useful AI tools. The question is whether their internal deployment strategy lets the good tools surface naturally or buries them under mandates that make workers resent AI assistance in general. That resentment is the hidden cost no adoption metric captures - once employees associate AI tools with busywork, getting them to engage with tools that actually help becomes much harder.