Shopify CEO: Most Companies Use AI Wrong—Machines Can't Take Responsibility

Shopify founder and CEO Tobi Lütke said in a Sept. 15, 2026 episode of The Knowledge Project podcast that most companies are using AI incorrectly, treating it as a way to avoid thinking, pile on features and shift responsibility. He described what Shopify internally calls "slop grenades": employees use AI to generate code or long emails without review and pass them to colleagues, offloading verification costs to others.

Lütke argued that in the AI era, slacking off no longer means underproducing but overproducing. The right approach is to keep a human-in-the-loop, using AI as a source of diverse perspectives, reasoning checks and creativity, while humans make final decisions and take responsibility. "Machines can't take responsibility," he said.

He also discussed River, Shopify's internal AI colleague. River has a name, avatar, memory, access to the company's code and systems, works in public Slack channels, and can submit pull requests. About 50% of Shopify's pull requests now come from conversations in company chat. River also runs a nightly "dreaming" process to review her conversations and improve herself.

Lütke said taste, intuition and judgment will become the most valuable human assets as AI spreads, and building great products depends on pruning and rebuilding, not endless feature addition.