Shopify CEO Reportedly Warns Staff Forwarding Unreviewed AI Work to Colleagues

Tobi Lütke has spent the past year pushing Shopify hard into AI. Now he’s calling out one of the messier side effects.

In a recent appearance on The Knowledge Project podcast, the Shopify CEO described a growing internal problem: employees generating emails, code and documents with AI, then passing the unfinished results along without really checking them. He has a name for it — “slop grenades.”

“We call those ‘slop grenades’ that people toss at each other,” Lütke said. “And that’s definitely a bad thing.”

The timing is notable. In 2025 Lütke made AI use a baseline expectation at the company. Teams were told to prove they couldn’t get the job done with AI before asking for more people or resources. AI proficiency even became part of performance reviews. Shopify’s own internal agent, River, now handles a huge share of production code pull requests — Lütke estimated it could be as much as half.

The tools work. The problem, he says, is what happens when people stop taking responsibility for the output.

One common example: someone prompts an AI to write a long email, hits send, and the recipient ends up feeding the whole thing into another AI just to figure out what it actually says. “Why did we invent decompression and recompression?” Lütke asked. “This is terrible.”

The same pattern shows up in code. An engineer tells the AI to go for it, barely reviews the resulting pull request, and tosses it to colleagues. The time saved by the first person becomes cleanup time for everyone downstream.

Lütke’s point is straightforward. Machines don’t take responsibility. When something breaks or a message lands wrong, a human still has to own it. AI is most useful, he argues, when it helps people think more clearly and write more concisely — not when it simply piles more unexamined work onto someone else’s plate.

The issue isn’t limited to Shopify. Researchers have started calling similar material “workslop” — polished-looking AI output that still requires significant cleanup. A recent survey found more than half of U.S. desk workers admitted to sending it, while many who received it spent hours each month fixing it.

Lütke remains firmly in the pro-AI camp. Shopify continues to lean heavily into the technology, and he has spoken enthusiastically about more advanced internal uses. But his latest comments mark a clear shift in tone. After pushing the company to adopt AI aggressively, he’s now warning that the real productivity gains only show up when people stay accountable for what the tools produce.

In other words, the grenade still has a pin — and someone has to decide whether to pull it.

Latest Posts

[democracy id="16"] [wp-shopify type="products" limit="5"]