A leaflet proofreading turned an afternoon into unpaid labor. As AI pushes errors to customers, companies face hidden costs in time, trust, and churn. The risk is a widening divide between good and bad jobs.
A summer visit to a castle garden in rural Normandy turned into an unpaid translation shift. The leaflet handed to a German-speaking couple had been run through AI. It translated “lookout point” as “point of view.” The woman at the counter confirmed it was AI. The couple spent an hour correcting mistakes, a task that once would have belonged to a translator.
The anecdote, shared on the Crooked Timber blog, illustrates a pattern spreading across industries. AI handles routine tasks well, but its errors often get pushed to customers, volunteers, or casual passersby. The work does not disappear. It is redistributed, and the cost is borne by the people on the receiving end.
The blog post recounts another scene: an elderly man at a postal office, desperate after a string of failed calls with an automated system. His battery ran out during an endless hold. The system not only failed to fix his mistake but added new ones. The service worker on the counter had to deal with his frustration. The redistribution of work can be brutal when systems break down.
For companies, the equation is not clean. AI reduces cost in one column but can add cost in others: longer resolution times, higher churn, damaged trust. A leaflet that needs proofreading by visitors is a minor example, but the same dynamics scale. A bank that routes customers through automated loops may save on call-center staff while alienating depositors.
The blog post argues that the reshuffling of tasks often strips jobs of creativity and agency. Translation, once a craft of judgment, risks becoming error-correction under time pressure. Algorithmic management can speed up workflows but also tighten control, leaving less autonomy for workers. The divide between good jobs and bad jobs is likely to widen.
The author notes a bigger risk: AI systems are being introduced under power relations that favor employers and large service providers over workers and customers. The ability to survey and control more areas of human activity is real. Some jobs may improve as boring elements are automated, but the number of low-quality, algorithmically managed gigs is likely to grow.
The essay ends with a call to action. It points to a paper by Jelena Belic and Kritika Maheshwari that argues universities should teach strategies of resistance alongside critical thinking. Not just how to work with AI, but how to push back when it degrades work or service.
Prepared with AlphaScala editorial tooling from the source reporting linked above. Indexable analysis may include a cited Alpha Score value. Publishing checks screen each story before release. Educational coverage, not personalized advice.