Shopify CEO Criticizes Employee AI Use as 'Slop Grenades'

Shopify co-founder and CEO Tobias Lütke has expressed concerns regarding the unexamined and often sloppy output generated by employees using artificial intelligence tools. Lütke stated in an interview on The Knowledge Project podcast on Tuesday that employees are producing unexamined emails and code with AI and not taking responsibility for the resulting poor quality. He characterized this output as "slop grenades" that employees "toss at each other," creating additional work and headaches for colleagues who must review and correct the AI-generated content. Lütke elaborated that while AI is intended to streamline tasks, such as synthesizing points in an email, its misuse can lead to lengthy, unreadable "missives" that waste time and require extensive review. He suggested that employees are allowing AI to perform the work without proper oversight or accountability, which undermines the intended benefits of AI integration.

This sentiment echoes a broader shift in how some tech leaders view AI adoption in the workplace. Previously, many tech leaders championed AI as a means to enhance employee potential and efficiency. Lütke himself had previously communicated to Shopify employees that using AI was a "baseline expectation," requiring them to demonstrate that they "cannot get what they want done using AI" before requesting additional resources. However, the current perspective highlights the challenges of managing AI output and ensuring its quality and accuracy.

Duolingo CEO Luis Von Ahn has also voiced similar reservations about unchecked AI implementation. Last year, Duolingo announced an "AI-first" strategy, which involved evaluating employees on AI usage, replacing human contractors with AI, and only increasing headcount if a team could not automate necessary tasks. However, in May, Von Ahn told Fast Company that while AI demos showed promise in writing, its creative output at scale did not match that of Duolingo's human employees. He noted that when scaling AI-generated content, such as writing "1,000 different stories for people to learn a language," a significant portion, approximately 20%, could be "pure slop." Von Ahn emphasized the need for caution to prevent such low-quality output from propagating when scaling operations with AI.

The phenomenon described by Lütke and Von Ahn has been termed "workslop" by researchers. This concept refers to AI-generated output that appears polished on the surface but ultimately hinders productivity due to its inherent flaws, inaccuracies, or lack of genuine insight. The core issue lies in the uncritical adoption and deployment of AI tools, where the perceived efficiency gains are offset by the increased burden of reviewing, correcting, and rectifying substandard AI-generated work. This trend suggests a growing awareness within the tech industry that AI integration requires careful management, robust quality control, and a clear understanding of its limitations to truly enhance, rather than detract from, overall productivity and output quality.