how to automate phoning it in with a.i.

Business are lamenting that workers use generative AI to turn in subpar work, despite giving them every incentive to do so.

A common refrain about modern work culture goes something like this. People do not want to spend the one life they have around others they don’t particularly like or care about so they can continue to barely afford to exist. Which checks out, given that nine in ten people worldwide reported being frustrated and disengaged at work all the way back in 2016, and while the numbers went up during the pandemic thanks to remote work and a thriving job market, they’re once again rapidly declining.

With this context, it makes perfect sense that with the explosion of chatbots and the now economically concerning hype around them, workers are now complaining about colleagues who are using LLMs to take phoning in their jobs to a whole new level in a phenomenon that Harvard Business Review calls workslop. They churn out reams of low quality, error-filled output which just generates more work for everyone else.

This also extends to the tech world, where programmers who originally went into the field of computer science now lament that they’re letting AI write too much code for them while also knowing that it’s making them worse coders and feeding into the AI hype cycle that is openly and actively hostile towards our profession.

Passion for the work and aspirations to solve ever more complex problems slammed headfirst into corporate release cycles and laser focus on outcomes and completing tasks to meet investor-demanded metrics, and the latter won. Since the business is not interested in how elegantly you solved a huge problem or how good your code is under the hood, why bother to come up with the best and most elegant solution if AI byte vomit will suffice?

We can apply the same logic to any work where AI generated content will do the job, even if that job isn’t all that good. If your talent and your work will not be appreciated or recognized, or earn you a promotion, autonomy, or more time off, why expand the mental bandwidth on it? Might as well outsource it to an app, especially as the boss keeps saying that your job will be automated once they figure out how.

Now, you could argue that this is cheating your employer and customers as they’re paying for your best effort, and technically, you’d be right. Bad output means your colleagues have to fix all the hallucinations and mistakes down the line. Bad code is an enormous liability as it can fail in subtle ways, create security holes that can leak sensitive data, make new features too complicated to add, and bugs harder to fix.

At the same time, thanks to over-consolidation, consumers have been begrudgingly accepting lower quality for higher prices, and thanks to a dysfunctional job market, companies can treat their employees worse and worse, especially if they fully buy in on adopting AI for mass automation. If customers now expect shoddiness and effort isn’t being consistently rewarded, this simple assumption quickly breaks down when met with the complexities of reality.

Considering all this, yes, we could get upset by workslop and do the usual American management thing in promising that the beatings will continue until morale improves. Or, we could try something different and ask why so many people are phoning in the work, whether the work they’re doing is really meaningful, and where else would the workslop creators better be suited to make real, tangible impact and gain a sense of purpose instead of continuing to massively misallocate our human capital.