Too Much Length: When More Becomes Less
Too Much Length: When More Becomes Less
Stanford researchers just coined a new term: workslop. It describes AI-generated work that looks professional but lacks substance. And it's costing companies millions.
AI has given us superpowers. Tools like Google's Deep Research and Perplexity can tackle research projects that used to take teams weeks. We can scale our output in ways that seemed impossible two years ago.
But there's a problem.
The Seductive Trap of Easy Content
40% of workers received workslop in the past month, according to a BetterUp and Stanford study of 1,150 workers. These employees spent nearly two hours cleaning up each piece of AI-generated work.
AI creates content that looks credible at first glance. You hit generate, copy, paste, send. You look productive.
Most of us aren't reading what we're sending. We treat AI like a magic content printer. Sometimes the output is good. Often, it isn't.
The Length Problem
AI loves words. Ask for a market analysis and you'll get something longer than War and Peace. We save hours by having AI write for us, then waste those same hours trying to digest what it produced.
Or worse, we use another AI to summarise the first AI's work. Machines talking to machines while humans disappear from the conversation.
Sometimes the most impactful insights come in the smallest packages.
The Real Cost
The Stanford research shows workslop costs organisations £6.6 million annually for every 10,000 workers.
The human cost runs deeper. When you send AI-generated fluff, you tell colleagues I couldn't be bothered to think about this properly. 53% of workers feel annoyed when they receive workslop.
Half view the sender as less creative and reliable.
Trust erodes. Quality drops.
What Other Research Shows
This isn't just one study. Multiple independent research projects paint a complex picture:
MIT researchers found that 95% of AI pilot programmes fail to deliver measurable returns, despite companies investing £22-29 billion in generative AI.
MIT Sloan research shows AI can improve performance by 40% when used within its capabilities, but performance drops 19% when used beyond them.
Federal Reserve Bank of St. Louis studies found 28% of workers use generative AI at work, with users reporting meaningful time savings.
Yet experienced software developers took 19% longer when using AI tools, despite expecting them to speed up their work.
Meanwhile, Gallup data shows AI workplace usage has doubled from 21% to 40% since 2023.
But only 16% of workers strongly agree their company's AI tools are useful.
We're scaling the wrong thing.
The Prompting Problem
Much of this stems from lazy prompting. We treat AI like a mind reader instead of the sophisticated tool it is. Good AI output needs constraints, context, and clear expectations.
Set word limits. Define your audience. Specify format. Give examples of good work.
Would you ask an intern to write something about sales with no guidance, then send whatever they produced to your CEO? Then why do it with AI?
The Path Forward
The solution isn't abandoning AI. It's using it thoughtfully.
The time AI saves you in content creation should go towards the most human skill: judgement. Reading. Editing. Asking Does this make sense? and Will this help the recipient?
Treat AI as your research assistant and first-draft writer. Not your ghostwriter and final editor. Let it help you think through problems and structure arguments. Then add the insight that only comes from experience.
AI should amplify your intelligence. Not replace it.
Stop Creating Workslop. Start Creating Value.
Most people know AI can help them work faster. What they don't know is how to avoid the workslop trap.
That's why I run AI Cook-a-Longs. Designed to take you from curious to confident in one hour, there are no presentations about AI's potential and no theory about the future of work. Just sixty minutes where you learn to prompt properly, edit ruthlessly, and turn AI into a thinking partner instead of a content printer.
Leave a comment or direct message me if you'd like to take part.