Don’t let AI slop grenades blow up your business

You know the email. It’s long, tidy, confident — and somehow says nothing. You read it twice, then realize the real work of figuring out what it means just landed on your desk.

Shopify’s CEO has a name for it, and this week it made headlines. I think the story deserves a closer look, because the problem is real, the fix is well understood, and it isn’t what most people reach for first.

What Shopify’s CEO actually said

In a Tuesday interview on The Knowledge Project podcast, Tobias Lütke said Shopify employees are producing unexamined emails and code with AI and not taking responsibility for the output. He called these slop grenades: work that gets tossed to a colleague, who now has to review it and work out what’s wrong. His example was an email that AI should have boiled down to its key points but instead ballooned into a long missive. Fortune has the full story.

Worth remembering: last year Lütke told employees that using AI is expected of everyone, and that they should show they can’t get something done with AI before asking for more resources.

I want to be fair to him. This isn’t a leader walking away from AI. It’s a leader describing what happens when a powerful tool meets busy people who don’t share a definition of “done.” Saying it out loud is the first step, and plenty of companies haven’t taken it.

Why slop happens (and why it’s rarely laziness)

Researchers call it workslop: polished-looking AI output that lacks the substance to move a task forward, so the recipient has to do the thinking the sender skipped. The Harvard Business Review team behind the term found that it says less about careless individuals and more about the conditions leaders create: vague directives to use powerful tools, handed to teams who are overburdened and don’t feel safe saying “I’m not sure.”

Fortune’s coverage of a BetterUp Labs and Stanford survey of 962 U.S. desk workers puts a price on it. 38% received workslop and estimated it cost them 3.4 hours a month to revise. More than half said they had sent some themselves, and that was more common where organizations encouraged AI use. Read that last part twice. When “use more AI” is the only instruction, volume is the only thing anyone can show.

The professions that got here first

Other fields have already had their slop-grenade moment, and they answered it with governance, not guesswork. Here’s what they did.

The fix in every field was the same: a named human, a check before handoff, and a note that AI was involved. That’s the whole governance playbook, and it fits on an index card.

Smart governance: six norms any team can adopt

Here’s the version I’d hand to a marketing, sales, or RevOps team tomorrow.

Where to start this week

Pick one recurring handoff — the campaign brief, the weekly pipeline report, the outbound sequence — and add one line to its template: AI used: [tool or none]. Checked: [what you verified].

That single line makes ownership visible, tells the reviewer how hard to look, and starts the conversation that matters: what would make you comfortable putting your name on this? The answers become your norms.