Shopify’s CEO: ‘Slop Grenades’ Highlight AI Output Overload As Workplace Challenge
Shopify CEO Coins Term 'Slop Grenades' as AI Output Overload Becomes a Workplace Problem
Shopify CEO Tobi Lütke has a name for unreviewed AI output dumped on unsuspecting coworkers: a slop grenade. He warns the new failure mode of lazy work is too much output, not too little.
The warning lands at a revealing moment. Shopify has aggressively pushed AI adoption across its workforce since April 2025, yet its own CEO is now flagging a cultural side effect that no adoption memo anticipated. As AI tools become standard equipment in the modern workplace, the hidden cost of over-generation is quietly shifting from the person holding the keyboard to the person sitting across the Slack channel.
What Lütke Actually Said
Lütke made his remarks on The Knowledge Project podcast, released September 15, 2026. Host Shane Parrish asked whether AI had made anything worse inside Shopify. Lütke responded around the 16-minute mark of the episode with a candid answer that cut against the usual AI enthusiasm.
"The failure case now of lazy work is not lack of output," Lütke said.
He offered two concrete examples to explain the problem. In the first, an employee asks an internal AI agent to make a code change, approves the resulting pull request without reading it closely, and hands the review burden to colleagues. In the second, someone uses a language model to expand a short point into a long email. The recipient then shortens it again using another model.
"So we call those 'slop grenades' that people toss at each other," Lütke said.
His advice on the email scenario was direct: use the model to make a point shorter, not longer. He also noted earlier in the interview that machines cannot take responsibility for work. People can. Lütke credited Shopify engineer Harry Brundage with coining the term and expressed hope it would catch on more broadly in a post on X.
Why This Framing Matters
The slop grenade concept reframes a conversation that has largely focused on AI's capacity to produce rather than its capacity to burden. Most AI productivity discussions centre on time saved by the person generating output. Lütke's observation forces a different question: how much time is being consumed by the person receiving it?
This distinction is important for managers and team leads evaluating AI adoption. Output volume is not the same as productivity. A tool that generates ten pages of content in seconds can still produce a net loss if it takes a colleague three hours to determine whether any of those pages are usable.
The AI Adoption Backdrop at Shopify
The slop grenade comments arrive roughly 17 months after Lütke made AI use a formal expectation inside the company. In April 2025, he circulated an internal memo stating that "reflexive AI usage is now a baseline expectation at Shopify." Search Engine Journal's Roger Montti covered that memo at the time.
The memo carried real weight. AI use would factor into peer and performance reviews. Any team requesting additional headcount would first need to demonstrate why AI could not handle the work instead.
By October 2025, a Shopify engineer reported that the company had reached universal adoption of AI code editors. Every team received unlimited access to top AI models and thousands of Cursor licenses.
On the podcast, Lütke described River, an internal AI agent operating inside Shopify's Slack channels that can open pull requests directly. He estimated that up to roughly half of Shopify's pull requests now originate from conversations in that company chat environment. He also described using a personal team of AI agents to debate both sides of difficult decisions he faces.
The scale of adoption makes Lütke's concern more significant, not less. Universal adoption does not tell you whether work was saved or simply transferred to a colleague downstream. For businesses exploring what deep AI integration looks like in practice, understanding how Shopify works as a business platform provides useful context for why operational efficiency sits at the centre of Lütke's thinking.
The Accountability Gap
What makes the slop grenade problem particularly difficult to address is that it is largely invisible in standard productivity metrics. A team might report higher output, faster turnaround times, and fewer blockers — while quietly accumulating review debt that never appears on a dashboard.
Lütke's core argument is one of accountability. Machines generate. People are responsible. When that distinction blurs, the cost does not disappear — it relocates. Understanding the risks and challenges of AI in business is essential context for any organisation navigating this shift, particularly those that have moved quickly toward universal adoption without building review and accountability structures to match.
A Broader Market for AI Cleanup Is Already Forming
Lütke's internal observation reflects a pattern showing up across the wider labour market. Data shared by the Guardian on September 2 pointed to rising freelance demand for individuals who correct AI-generated work.
On Freelancer.com, listings tagged with phrases like "correct AI" and "AI hallucination" grew by 87% from August 2025 to June 2026, reaching 10,760 listings worldwide. Upwork reported a 70% year-over-year surge in AI remediation gigs. Fiverr searches for "AI cleanup" services increased more than 20-fold between 2023 and 2026.
Inside companies, the cost registers differently. A BetterUp survey conducted in September 2025 with the Stanford Social Media Lab asked more than 1,000 full-time U.S. desk workers about their experiences. About four in ten respondents said they had encountered "workslop" in the past month. They described it as AI-generated work that appears polished but fails to advance actual tasks. Each instance took nearly two hours to resolve on average. The data is self-reported.
For content and marketing teams using AI for initial drafts, those two hours represent a real line item in the production budget, even if it never appears on an invoice.
How Long Will the Cleanup Economy Last?
Freelancers surveyed in the Guardian report disagreed on the timeline. One writer expected much of the cleanup demand to dry up within five to ten years as models improve. A designer predicted AI would match her logo work within a couple of years.
The honest answer is that nobody knows. What is clear is that the gap between AI-generated output and output that is genuinely useful to its recipient is currently large enough to support a measurable freelance economy built entirely around bridging it. Whether that gap closes through better models, better prompting habits, or stronger workplace accountability norms — or some combination of all three — remains an open question.
For smaller organisations still assessing how and where to integrate AI, the cleanup economy offers a useful signal. The practical guide to AI for small businesses is worth reviewing before committing to high-volume AI generation workflows without the review infrastructure to support them.
What This Means for Teams Using AI at Work
The slop grenade problem is ultimately a workflow and accountability problem, not a technology problem. Lütke's framing offers a practical lens for any team navigating heavy AI adoption.
Three conclusions follow directly from the evidence in this story:
Reviewing AI output before passing it along is part of the actual cost of using the tool. Ignoring that cost does not eliminate it — it transfers it to whoever receives the output next. That transfer is invisible in most productivity tracking but very visible to the person on the receiving end.
Organisations measuring AI adoption by output volume may be tracking the wrong thing. The more meaningful metric is how much downstream review time that output is generating. A team producing twice as much content is not more productive if a separate team now spends twice as long verifying it.
The most effective use of a language model in communication is often compression, not expansion. Lütke's advice is direct: use the model to make a point shorter. Shorter, clearer output reviewed by the person who generated it is worth considerably more than a long document that arrives as an unexpected burden in a colleague's inbox.
The slop grenade is, in the end, a metaphor for misallocated responsibility. AI generates. People decide. Teams that maintain that distinction clearly — regardless of how deeply they adopt AI tooling — are better positioned than those that let the line blur in the name of speed. For a deeper look at how these dynamics are reshaping workplaces, MIT Sloan Management Review's ongoing AI and work coverage offers rigorous, research-grounded perspective.
