According to Fortune, about 40% of U.S. desk workers encounter AI-generated “workslop” each month, with each incident taking an average of two hours to resolve. This costs companies approximately $186 per employee monthly, adding up to $9 million annually for a 10,000-person organization. MIT researcher Michael Schrage predicts workslop will become a major governance challenge, with senior management demanding “workslop metrics” similar to quality metrics. He anticipates companies will fight AI with AI by training models like ChatGPT to detect and filter slop automatically. Schrage also expects organizations will increasingly require employees to share their prompts as proof of genuine work, making transparency non-optional.
The real cost of looking busy
Here’s the thing about workslop – it’s not just annoying, it’s expensive. When you do the math on those numbers, you’re looking at nearly a full workday per month per employee wasted on cleaning up AI-generated nonsense. And that’s just the direct time cost. What about the opportunity cost of people not doing actual thinking? The BetterUp research shows this is becoming a systemic problem, not just occasional frustration.
Think about it – we’ve all seen those beautifully formatted but utterly useless AI reports. They look polished enough that someone probably spent time on them, but they don’t actually move anything forward. It’s the digital equivalent of rearranging deck chairs on the Titanic. The scary part? This is happening while companies are pushing AI adoption harder than ever.
Fighting fire with fire
Schrage’s solution is both clever and inevitable: use AI to detect AI slop. Basically, you train your corporate LLMs to recognize when other LLMs are phoning it in. It’s like having a plagiarism detector, but for lazy thinking rather than copied text. And honestly, this makes perfect sense – if AI can generate convincing-looking garbage, AI should be able to spot it too.
But here’s where it gets really interesting. Schrage suggests that slop detection might become an “underground” tool. Imagine managers quietly running their team’s work through slop detectors before meetings. That creates a whole new layer of workplace dynamics. Will employees know they’re being monitored? Will there be slop detection standards? This could become the new quality control frontier.
Show your work means show your prompts
Schrage’s approach at MIT is revealing – he’s basically given up on fighting AI use and instead demands transparency. “Show me how you’re prompting your work” is becoming the new “show your work.” And he’s probably right that this will spread to corporate America. Think about it – your prompt history could soon matter as much as your performance reviews.
That’s a radical shift. Performance reviews measure outcomes, but prompts reveal process and thinking. In an AI-heavy workplace, how you approach problems might become more important than the solutions you deliver. After all, if the AI can deliver the solution, what are you actually bringing to the table? Your judgment in guiding the AI, apparently.
The compliance dilemma
For companies worried about feeding proprietary data to external AI models, Schrage offers a clever workaround: analyze competitors instead. Use publicly available data like earnings calls and filings. It’s a way to get AI insights without the security risks. This approach actually aligns with broader trends in competitive intelligence and market analysis.
The bigger picture here is that we’re entering an era where AI literacy becomes as important as any other professional skill. As EY’s workforce research shows, the nature of work is fundamentally changing. And with economic pressures mounting, companies can’t afford to pay people to clean up AI messes. The message is clear: either learn to use AI properly, or prepare for some uncomfortable conversations about your “professional judgment.”
