Your boss can tell
You pasted the answer out of a chat window, smoothed one sentence, and hit send. It read fine to you. It did not read fine to the person who received it, and the research now exists to put numbers on exactly how not-fine.
In late 2025, Harvard Business Review published a study from BetterUp Labs and Stanford researchers that gave the phenomenon a name: workslop, AI-generated output dressed up as finished work. The numbers were blunt. Roughly four in ten of the desk workers surveyed had received workslop from a colleague within the past month. Each incident cost the receiver around two hours of cleanup. And the finding that should stop you mid-paste: over forty percent said they trusted the sender less afterward, and about a third wanted to work with that person less. Not the tool. The person.
Notice what's actually being measured there. Nobody ran a detector. Nobody filed a complaint. The receivers simply read the thing, recognized that the thinking had been skipped, and quietly repriced their colleague. That's the mechanism: workslop doesn't eliminate the work, it delegates the work upward or sideways to whoever receives it. They have to reconstruct the context you didn't provide and make the decisions you didn't make. People notice when they're handed their own homework back. They remember who handed it to them.
The asymmetry is what makes this expensive. The time you saved was measured in minutes. The trust you spent recovers on a scale of months, if it recovers, because trust between colleagues doesn't get re-earned by one good memo. It gets re-earned by a streak of them, under scrutiny you didn't have before.
The fix is not abstinence. The machine is genuinely useful, and pretending otherwise is its own kind of career mistake. The fix is refusing to ship its default voice under your name. At minimum, that means reading the draft once before it leaves you, which catches the pasted headers, the chatbot sign-offs, the sentences no human would say in a meeting. One step better: run it through the Slop Check and fix the exact lines it flags, ten minutes, done. And the durable version: stop editing yourself back in after the fact and put yourself in before the draft exists, which is the job the Voice Kit was built for. Your rhythm and your vocabulary go in first, so what comes out already sounds like the person whose name is on it.
The machine didn't spend your credibility. The send button did.
Before the next send: the Slop Check flags the tells in your browser, free. To make drafts start in your voice instead of needing rescue, that's the Voice Kit, $19, once.