Copy article

New Study Reveals $9 Million Workslop Tax: When AI Shortcuts Create More Work, Not Less

ended 26. September 2025

Fresh research conducted in September 2025 by Stanford Social Media Lab and BetterUp Labs exposes a hidden productivity killer. “Workslop”, AI-generated content that looks polished but lacks substance, forces frustrated colleagues to decode, correct, or completely redo the work of others. 

Their survey of 1,150 full-time workers found 40% received workslop in the past month, with staff estimating that 15% of all content they receive qualifies as low-effort, unhelpful, AI-generated work. The study revealed that workslop is a particular scourge in the professional services and technology sectors.

It's not just time that's wasted. Since each incident costs nearly two hours to resolve, it causes a monthly productivity tax of around $186 per head. Scale that up for a 10,000-person organisation, and that's $9 million annually in lost productivity.

Compounding the problem is the increase in workplace stress and crumbling team morale. One project manager explains: "Since [the workslop] was provided by my supervisor, I felt uncomfortable confronting her about its poor quality. So instead, I had to take on effort to do something that should have been her responsibility, which got in the way of my other ongoing projects."

Is the ‘AI arms race’ creating a generation of workplace freeloaders who think hitting "generate" counts as doing the work? Beautiful-looking deliverables that are fundamentally broken, dumped on colleagues who can't push back without looking difficult isn't sustainable. 

KPMG's 2025 Trust in AI report reveals the governance vacuum fuelling this chaos. Employees reported only 34% of organisations have policies guiding generative AI use, while 48% of staff confessed they have uploaded sensitive company information, such as financial, sales, or customer information, into public AI tools. Half feel pressured to use AI or risk being left behind, yet only one in two reports receiving any training in responsible AI use. 

Employers urgently need to consider workflow AI guardrails, because if they don't a toxic culture will develop organically.

  • What human-oversight mechanisms are needed to prevent workslop before it reaches colleagues and/or customers? 
  • How can organisations develop robust generative AI workflows to maximise the productivity and quality rewards and minimise the risks?
  • How should a company prevent staff uploading sensitive information to their personal free accounts on platforms?
  • How should a disgruntled employee raise the subject of workslop with their colleagues and superiors?

     

5 responses from the Newspage community

Copy all

Star Quote
Copy

Colette MasonVerified
AI Ethics Consultant at Clever Clogs AI
Human-centred AI means purposeful and productive human-AI partnerships. Workslop snuffs out that partnership when staff confuse 'assistance' with 'replacement'. When presented with poor quality work, suggest the need to improve processes, rather than specific pieces, so individuals don't feel singled out. Train staff to use AI responsibly. Make them aware of the risks of using public platforms for confidential data. Better still, give them dummy data to experiment with, and a secure AI environment to play in, rather than leaking sensitive customer data on live public platforms. Re-evaluate core processes and assess where AI adds value (speed, rigorous analysis, ideation) and where it adds risk (overlooking hallucination, reputational damage, rework, frustration and burnout) and amend them accordingly.
Copy

Editor
Samuel Mather-HolgateVerified
Managing Director & IFA at Mather and Murray Financial
AI is a tool and needs to be used like other language tools, like a dictionary. If used well, as inspiration, it works great. However, for those who fall into the trap of relying on it will see their work standards decline and service fade. AI can increase productivity and standards, but human intervention remains key in setting guardrails and adding soul to the machine.
Copy

Premium
Debbie PorterVerified
Managing Director at Destination Digital Marketing
The KPMG's 2025 Trust in AI report seems to have employees pinning incorrect AI use on the employer. As an emerging technology, those ‘at the top’ are also still wrestling with what generative AI is and what its capabilities are. If 48% of staff are confessing they have uploaded sensitive company information, such as financial, sales, or customer information, into public AI tools then the responsibility for not doing this would lie with the employee, just purely on a sense-check. Not sharing sensitive information with your competitors and other third parties is already covered in contracts and staff handbooks, so some of this chaos is being caused by employees misstepping in the first place without the knowledge of their employers. Blindly producing workslop should be dealt with like any other workslop - through HR and supervisory processes. No matter how something got produced, if the quality is poor, then the quality is poor.
Copy

Keith BuddenVerified
Managing Director at Ensurety
The unauthorised use of AI is I think potentially one of the biggest dangers to personal data - I recently had a client who advertised a vacancy. They were expecting around 40 applicants but instead got 200. They decided to use ChatGPT to scan all of the applications, weigh the contents against select criteria and filter out the 20 'best fit' applications. So far so good, except a) they were still using the 'free' version of ChatGPT (so anything they entered fed the main ChatGPT LLM) and b) they did not redact any of the personal information before processing the 200 CVs. This was a serious issue, especially that unlike a normal database where you could simply delete this data, deleting this data was now not an option.

So should you use AI in your business, probably yes, BUT it must be used with proper controls and governance - that way is the path to paradise, all other paths lead into the fire pit of hell.
Copy

Ben Foster0
CEO at The SEO Works
AI has the ability to produce work that on the surface looks polished, believeable and professional. But when you dig under the surface you start to realise it holds little value, uses lots of words and repetition, and had no unique insight (potentially with a few hallucinations thrown in for measure). On the flipside, if used well, it can create powerful insights from huge datasets and fantastic outputs.

I think we need to accept that if employees are being encouraged to use AI then this will happen. Lazy people will always look for a shortcut - so before AI, these employees were probably still producing low-effort work.

The message is that human oversight is ALWAYS needed to get the best out of AI. If someone receives workslop then they have the right to call it out, the same as if they received a poor quality piece of work before AI existed.

This "personal ownership" of quality should be combined with robust training so that employees are aware of the good and bad points of AI.