The Rise of Workslop
Around here at M2 we’re pretty pro AI, we have a whole event centering around helping businesses how to prepare for it and integrate it with their business. Of course with any tool, proper application is key. When you have an AI hammer that seemingly promises to fix everything, the entire planet starts to look like a nail.
A report from MIT has found that of the $30-40 billion that has been poured into integrating AI tech into business, 95% of businesses are seeing no measurable return, while 5% are extracting “millions in value”. The report found that “This divide does not seem to be driven by model quality or regulation, but seems to be determined by approach.”
Enterprise grade AI systems have been failing to gain the ground they were expecting, with most businesses instead rolling with the far easier to implement AI that enhances individual performance. That way employees can use their discretion for implementation and smoothing out the roadbumps they face on a granular level.
This isn’t the end of the story though. What we’re starting to see is an emergence of poor execution undermining businesses at the most basic level. It Turns out poor execution can cost you more than you’d think. A new report from Stanford and BetterUp Labs has found that employees are becoming increasingly stymied by what is being described as workslop. This is AI generated text that has been barely moderated by the user and only serving to increase the downstream workload for the receiver. 40% of those surveyed have reported that they have received content masquerading as competent, but making no sense at all in the real world. While most of this useless content comes from peers, a certain amount comes from up high from leadership handing nonsense down the chain. This leaves the receiver with the task of checking anything that’s been generated is even correct, and rewriting the majority of it anyway, or going back to whoever generated it and getting them to close Minesweeper and do it properly this time.
The run on effect was that users responsible for workslop are the emotional crumplezone for poor AI output. Receivers judged them as being uncreative (54%), less trustworthy (42%), and less intelligent (37%).
In time lost it was estimated that it cost an individual employee around US$186 a month to clean up after AI. While we could create some scary numbers by extrapolating out to the largest possible workforce size, not every single person is going to be using AI in these businesses. Suffice to say you could be paying 16 times the cost of a monthly Chat GPT subscription to fix what it’s giving you depending on how much your personal time is worth.