Is AI Making it Easier for Workers to Cheat?

A recent comic strip shows two different office workers in two different cubicles, talking about the great benefits of artificial intelligence on their working lives—and on the company’s productivity as a whole.

One is saying to a colleague, with a certain level of delight, that “AI turns this single bullet point into a long email that I can pretend I wrote.”

In the cubicle on the far side of the office, another worker is telling a colleague, with similar delight, that “AI can make a single bullet point out of this long email that I can pretend to read.”

If you’re a casual consumer of the business press, you might have been led to believe that AI is going to supercharge our global economy and make all of us 25-50 percent more productive at whatever we do in our business lives.  But a recent article asks whether that is actually true for you, the reader of this (and her) message.  

If not, then is it possible that everybody else is riding the AI productivity wave and you’re the only one left behind?  Or, alternatively, is AI riding a hype wave?

The author of the skeptical report says that she was asked to edit a long document that was (as a lot of long documents are in first draft) badly written.  She spent three hours cleaning up the text, only to discover that the result was a pile of rhetorical fluff—and that the document had been written by one of the AI models.  Her point is that it is nearly impossible to prevent humans from cheating when they think they can get away with it, and AI has greatly lowered the barriers to cheating.

Cheating is a strong word, but if the use case of AI is to allow workers to pretend to write emails, memos and documents, and the recipients to pretend to be reading them, then business are introducing a problem into their midst.  Corporate strategists who are stumped can turn to their AI ‘assistant’ to come up with the next major move or innovation.  Executives can rely on a summary of a contract instead of reading the details.  A detailed chart created by hand might pass through layers of corporate oversight where nobody ever bothers to understand the logic behind it—or catch the flaw in the data that leads to a dangerous conclusion.

The most often-cited productivity increase is in customer service, where the data suggests that a staff person can handle anywhere from 15-30 percent more customers per hour using AI tools.  But corporate customer service departments reported similar productivity gains when they created phone trees that never actually led to an actual person, but instead would ultimately direct distraught customers to the company website’s FAQ section.  Productivity was up, but customer service took a nosedive.  Are we seeing the same story played out all over again?

Cheating looks like productivity in the short term.  Introducing AI functionality into a corporate setting might actually encourage people to complete their tasks by copying the output of a large language model that doesn’t actually think.  

The point here is an evergreen suggestion: tune out the hype when making decisions, especially investment decisions.  Hype can be roughly translated as overheated marketing, which is never a good window into the underlying reality.

Sources:

https://www.machinesonpaper.com/ai-and-the-illusion-of-productivity/ 

https://www.knapsack.ai/blog/is-ai-overhyped