AI Workslop: How to identify low-quality AI output at work

In a way, spotting workslop isn’t hard once you know what to look for. Basically, if something seems too polished, given what you know about that colleague or client, it was probably written by AI. Simon Kemp, founder of Kepios, a Singapore-based advisory company, explains: “There are certain people in organisations I’ve worked with over the past year who have shared things with me, and I think, ‘This is not how you speak or behave in real life.’ So the main question to ask is: does it match the person?”

Simon notes that in service industries in particular — anywhere writing is important — AI has become an integral part of daily activity. “Especially when it comes to management communication, like emails and things like that, it’s inescapable,” he says.

Spotting workslop — according to its creator

The best way to spot workslop is really knowing the person you’re interacting with, making sure the content ultimately makes sense and has an actual point to it instead of going round in circles. BUT, if you still aren’t sure in the end, here’s what ChatGPT has to say about it:

1. Overly tidy structure and “template” phrasing

• AI often produces writing that is: perfectly organized in neat sections
• Full of transitions like “In conclusion,” “Furthermore,” “It’s important to note that…”
  Stacked with generic clichés (“rapidly evolving landscape,” “crucial to understand,” “leveraging the power of…”)
• Humans are messier. They break patterns more naturally.

2. Vague claims with no concrete detail

• Avoids specifics unless prompted
• Stays general to avoid being wrong

3. Repetition & circular points

• Restates the same idea in slightly different words
• Loops back to earlier points without adding anything
• Often uses recycled concepts like  “collaboration,” “innovation,” “efficiency,” “importance,” etc.

4. Emotionally flat or unnaturally upbeat tone

• A dead giveaway is the overly enthusiastic, polite, or balanced style, such as “This offers a wide range of exciting possibilities.”
• Or an emotional tone that doesn’t match the topic.
• Humans are usually more opinionated, sarcastic, or idiosyncratic.

5. Lack of strong personal voice

• Uses consistent diction across very different contexts
• Versus humans who show unique word choices, personality, opinions and small inconsistencies

6. Weird factual errors mixed with confidence

• Classic AI errors include mixing up dates, names, details
• Makes confident but wrong claims and misattributes quotes
• Humans get things wrong too, but not in the pattern LLMs do.

7. “Smooth but empty” paragraphs

• A lot of AI text looks polished, but says nothing. If you can remove an entire paragraph and nothing is lost, that’s a sign.

Do these tips help you spot AI workslop better? To find our more on how AI workslop has shaped the workplace, and the importance of doing our due dilligence, read part one of this two-part feature here.