AI News Hub Logo

AI News Hub

What Looks Like Busywork Is Mostly Rent

DEV Community
Michael Tuszynski

Carl Frey's recent NYT piece argues that AI's real impact isn't job replacement — it's labor transfer from worker to consumer. We become our own travel agent, accountant, exterminator, doctor. The work doesn't disappear; it moves out of the labor statistics and into our evenings. Productivity climbs, corporate profits climb, individuals feel overburdened. The observation is correct. The framing is wrong. And the receipt that demolishes the framing is sitting inside Frey's own piece. Frey cites a family that used Claude to cut a hospital bill from $195,000 to under $33,000 — over $162,000 in coding errors and duplicative charges. He presents this as a "tangible benefit" of self-service, then immediately pivots to "however, self-service does not automatically reproduce a professional's judgment." That pivot is doing extraordinary work. The professional system here — the hospital's revenue-cycle department, the medical billing specialists, the insurance reviewers, the patient advocates — exists specifically to catch coding errors and duplicative charges. Those roles were billed for. They were paid. They simply weren't doing the work. The "burden" of the family doing that audit themselves wasn't AI making them busier. It was AI revealing that the prior arrangement was charging $162,000 for oversight that wasn't happening. The professional wasn't displaced. The professional was already absent — the bill just looked like they were there. This isn't a one-off. Industry estimates put error rates in medical bills as high as 80%, with average mistakes on $10,000+ bills running $1,300. The system supposed to catch them — the same system Frey mourns as professional expertise — has been failing silently for decades. AI didn't transfer the auditing burden to consumers. It revealed no one was auditing in the first place. The Frey argument conflates two fundamentally different categories of professional work: Access-rents — work where the intermediary's value was gatekeeping the inconvenience of access, not delivering judgment. Travel agents reading flight schedules. Tax preparers running TurboTax-style forms. Stock brokers placing trades. Bank tellers handling deposits. The intermediary added little beyond being a required step. Killing them is liberation, not burden. Integrated expertise — work where the professional integrated context and judgment the consumer couldn't reach. Differential diagnosis on ambiguous symptoms. Real tax strategy across multiple business entities. Trial strategy under specific judges. These require an expert who tells you what to ask, not just answers your question. Killing these is real risk transfer. AI eats both. Frey treats them as one phenomenon — "busywork landing on us" — and concludes we're overburdened. The conflation matters because the policy and product implications are opposite: accelerate the killing of access-rents, protect integrated expertise. The hospital billing example is squarely the first category. The patient wasn't replacing a clinical judgment with Claude. They were replacing the administrative oversight layer that was billed for but not delivered. That isn't thinner expertise. That's a system finally getting audited. Frey raises self-service travel as a historical parallel. He's right that the work moved. He's wrong that it transferred a burden. Travel agent employment dropped 60-80% between 2000 and 2020 as Expedia, Kayak, and Booking.com took over leisure bookings. Airlines eliminated most commissions to agents in 2002, removing the revenue model. Consumers did not drown in travel-planning busywork. The bookings that took 90 minutes with an agent (call, hold music, faxed itinerary, callback the next day) take 5 minutes online. The work didn't transfer to the consumer — most of it disappeared, because most of it was the friction of going through a human intermediary in the first place. The agents who survived the collapse were the ones doing real integrated work: complex multi-leg corporate travel, custom itineraries for high-end leisure, expertise on visa requirements and disruption handling. The access-rent agents disappeared. The integrated-expertise agents didn't. The market separated category A from category B without any policy guidance. Consumers benefited. Frey leans on "opportunity cost neglect" — the documented tendency to overlook the value of time we give up when self-serving. He's right that we miss it. The inverse error is also documented and larger in dollar terms: we routinely overpay for professional services that don't deliver judgment over what an AI tool gives free. The $300 accountant who beats a free AI tool by $30 on your return is a net $270 loss. The lawyer who charges $400 to fill out a generic LLC formation is a net $380 loss. The travel agent who books the same flight you'd have found is a net commission loss. Self-service neglect cuts the consumer one way; expertise neglect cuts them the other. Frey only counts one direction. The shift Frey describes is real, but the right read is closer to this: the post-WWII service economy normalized paying intermediaries to gatekeep our own affairs. Filing taxes was always something we could do; the IRS publishes the forms. Booking travel was always something we could do; the airlines publish their schedules. Disputing medical bills was always something we could do; the line items are itemized. We outsourced these tasks because the access was inconvenient and the time cost was real. AI dropped the access cost to near-zero and the time cost to minutes. What feels like busywork is mostly the recovery of agency over our own affairs. We're not drowning. We're doing things we could always do, finally without paying for the privilege. The argument for protecting integrated expertise still stands. The pediatrician who notices the symptom you didn't think to mention, the tax strategist who sees the structure across years of returns, the lawyer who reads the judge before the brief — these are real and AI is closer to replacing them than most professionals admit, but not there yet. Those roles deserve defense. The argument for protecting the medical billing specialist who wasn't auditing your bill, the travel agent who read the same schedule you can see, the accountant who clicked through TurboTax for you — that argument is over. It ended the day Claude found $162,000 in errors a paid system missed. As I argued yesterday, workflow-automation-as-pure-interface businesses are getting eaten first. The same logic applies to the human version of those businesses. What Frey calls busywork is mostly rent we finally stopped paying.