Shopify CEO Tobias Lütke says employees are using AI in ways that leave more work for everyone else.
In a Tuesday interview on The Knowledge Project podcast, Lütke said workers are producing unexamined emails and code with AI and not taking responsibility for the result. 'We call those 'slop grenades' that people toss at each other,' he said. 'That’s definitely a bad thing.'
Lütke said AI should help synthesize ideas, not turn simple messages into long drafts that colleagues must fix. 'You don’t really read it, and now it has to be reviewed by your colleagues,' he said. 'You’re just letting AI do the work for you.'
The remarks mark a notable shift from last year, when Lütke told employees that using AI was 'a baseline expectation' and that they should first prove they 'cannot get what they want done using AI' before asking for more resources.
The broader problem has a name now: 'workslop.' BetterUp Labs and Stanford’s Social Media Lab surveyed 962 American full-time desk workers this year and found that 52.7% reported sending workslop to colleagues. The survey said it was more common in organizations that encouraged AI use.
More than a third, or 38%, reported receiving workslop and said it took 3.4 hours a month on average to revise it. Respondents described it as polished-looking AI output that seems legitimate on the surface but still needs revision.
Examples included well-structured emails with broken links or code that was more complicated than it needed to be. Workers also said suspicion of workslop hurt relationships: more than a third, 36%, said they wanted to avoid working with colleagues who sent it in the future.
The survey estimated the cleanup time at $186 per month for single employees and as much as $9 million a year in lost productivity for an organization with 10,000 people.
Duolingo CEO Luis Von Ahn has also softened his stance. He said in May that he had gotten carried away by AI demoing well in writing, but said it does not match the creativity of Duolingo’s people when it scales.
The market has already voted on the problem: AI can raise output, but only if management insists on quality, accountability and real productivity. Otherwise, the taxpayer-adjacent lesson applies in every office: someone always pays for the cleanup.
