The 10-Second Version (AI Summary)
What happened: Companies that encouraged AI use are setting rules against "workslop," the AI-generated messages and documents filling their inboxes, and at least one employee has been fired over it.
Why it matters: Once someone feels they're getting slop from you, they stop listening, whether it's a coworker or a customer.
What to do: Set clear rules for where your team uses AI, and keep customer communication in their own voice.
AI-generated and reviewed by the ASOTU team.
AI is everywhere now, and most of the time it's obvious.
We've mostly stopped taking note of it. The thing worth flagging is false authorship: people dropping whole ebooks on Amazon as fast as they can ask Claude for chapters. Books that not even the author has read were not on our 2026 bingo card.
At work, the bar is simpler. Getting caught is the embarrassing part. We dug into why on today's episode of the Automotive State of the Union.
Twill CEO Michelle Volberg gave her 10-person team an enterprise Claude subscription, then asked them to stop talking to her like a robot.
Volberg says the Slack messages and emails coming from her staff started reading like pitches from clueless outside consultants, with perfect punctuation, missing client context and the occasional hallucination. Her fix was to keep AI for brainstorming and require people's own voices for internal messages.
She isn't alone.
Company | Its AI writing rule |
|---|---|
Twill | AI for brainstorming, your own voice for internal messages |
Clay | Stand behind every idea in every sentence, and a document should take longer to write than to read |
Spellbook | Customer messages are always handwritten, and five scrappy bullets beat five polished pages |
Zocdoc | Never ship your slop, and be transparent when a bot is talking to a human |
FischTank PR | No AI in anything sent to reporters or clients, checked with an AI detector |
FischTank's founder fired one employee who kept using AI for off-limits work after a confrontation and a warning.
How common is workslop in 2026?
In this year's Stanford Social Media Lab and BetterUp Labs survey, 52.7% of desk workers admitted to sending workslop. The share who received it dipped from 40% to 38%, and those workers estimated it cost them 3.4 hours a month to fix.
"Workslop is predicted by the conditions more than it is in individual behavior."
Kate Niederhoffer, chief scientist, BetterUp Labs
Once someone thinks you're sending them slop, they stop listening.
The problem started when a report took real time to make and we read it accordingly. Now people generate things they've only skimmed and put them on someone else's desk. That person has AI respond, and soon everyone is passing around poorly thought-out work with far too many words in it.
One fix Kyle picked up from a software engineer he works with: tell the AI to be terse. It cuts output to what a person would actually type, and it makes review faster.
The trust cost is real, too. Paul recently hit send by accident on an AI-drafted email before he'd written the personal intro he planned to type himself. A follow-up fixed the error, but the doubt it creates is harder to fix.
For dealers, that applies to every follow-up text and service reminder your team sends.
"[W]hatever you're doing with AI, people... hear and they see and they feel your heart based on how you treat them."
