AI Might Help Doctors Be More Efficient and Lower Medical Costs
A Northwestern University clinical study found that generative AI sped up radiology documentation by 15.5 percent.
At September's National Conservatism Conference, Sen. Josh Hawley (R–Mo.) argued AI "threatens the common man's liberty" and that "only humans should advise on critical medical treatments." Yet AI promises to enhance the human experience by reducing the price of critical services like health care.
The price of medical care has outpaced the consumer price index by 35 percentage points since 2000. The real culprit isn't administrative bloat but, as George Mason University economics professor Alex Tabarrok and his Why Are the Prices So Damn High? co-author Eric Helland explain, the Baumol effect. As other sectors become more efficient, the relative cost of slow-growth sectors (such as health care and education) increases. In other words, doctors haven't gotten "faster" at healing people, so the relative price of their time climbs.
Fortunately, AI tools can make doctors more efficient. A Northwestern University clinical study found that generative AI sped up radiology documentation by 15.5 percent. Even modest boosts like that could reduce overall medical costs.
Beyond health care, AI's impact on bureaucracy looks even more promising. A 2024 study by the Alan Turing Institute found that 84 percent of the U.K. government's 143 million routine processes are "highly automatable." Researchers estimate that saving an average of one minute per transaction would free up "the equivalent of approximately 2 million working hours per year." A follow-up paper published in June 2025 showed where AI could help most: Email tasks (estimated at 30 minutes per public sector worker per day) could be cut by 70 percent, and updating records or databases (estimated at 15 minutes per day) could be nearly eliminated.
To be sure, time saved doesn't always equal more work done. A May 2025 University of Chicago study found that AI chatbots shaved off about 2.8 percent of workers' time but "had minimal impact on productivity and labor market outcomes to date." Daron Acemoglu, professor of economics at the Massachusetts Institute of Technology, projects AI will increase productivity by only 0.55 percent over the next decade.
Even if AI doesn't spark a productivity boom, it can still reduce the daily grind. In health care, AI assistance means faster diagnosis of potentially life-threatening conditions. It's not a panacea, but it is progress, and Congress shouldn't stand in the way.
This article originally appeared in print under the headline "AI vs. Paperwork."
Editor's Note: As of February 29, 2024, commenting privileges on reason.com posts are limited to Reason Plus subscribers. Past commenters are grandfathered in for a temporary period. Subscribe here to preserve your ability to comment. Your Reason Plus subscription also gives you an ad-free version of reason.com, along with full access to the digital edition and archives of Reason magazine. We request that comments be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of reason.com or Reason Foundation. We reserve the right to delete any comment and ban commenters for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please to post comments
Getting government out of the health insurance and reimbursement industry should also increase efficiency and lower costs.
I bet prices would drop in half within a few months, and drop in half again within a few years. People just don't realize how much all those little bureaucratic meddlings add up. While rockets and health insurance are not directly comparable, this comparison of NASA and SpaceX is scary just for how incredibly incompetent government is.
https://caseyhandmer.wordpress.com/2025/10/31/nasas-orion-space-capsule-is-flaming-garbage/
NASA says each SLS launch will take 4 years to build, cost $5 billion, is 100% expendable, and it can't even do what Saturn V and Apollo did 55 years ago! They have spent more every year for the past 19 years than SpaceX has spent on its entire Starship development. Yet Congress keeps dumping money into it, in spite of having said the program will terminate after two more $5 billion missions.
Government fucks everything up.
I used to patronize concierge service for healthcare. The costs were known, upfront, and low. The few times I used insurance, there were more people in that office completing govt and insurance forms than providing healthcare services. The costs were unknown upfront and the final tally was considerably higher. But rent-seeking and subsidizing lazy folks was the goal and that has been achieved.
Last time I got my own insurance, about 20 years ago, I had catastrophic coverage. I don't remember the premiums at all well; maybe $200 a month. I paid everything under $2400, we split costs up to $3200, and they paid everything or most of everything over $3200. No way that would be possible now, even if everything were doubled for inflation.
I've asked several doctors how much less they could charge without insurance paperwork, and the answers have been 1/2 to 2/3. I'd bet that most of the insurance cost is from mandated government complexity, that if insurance and health care were in a free market, 90% of that complexity would disappear.
Yes it will because it will be more efficient to deny people health care and less expensive because many people will have no coverage.
And just what is your plan, Dr. Retard?
You take care of your health with your money to protect me, and vice versa.
What is your fair share of what someone else has earned?
"The price of medical care has outpaced the consumer price index by 35 percentage points since 2000."
Funny thing, so has "education".
It seems the more the government is involved, the faster costs rise.
^^
"AI Might Help Doctors Be More Efficient and Lower Medical Costs"
And pigs might fly.
AI is terrible at math and routinely hallucinates answers to technical questions when precision is critical.
I asked Grok how much of FDR's New Deal legislation would have passed if it took 2/3 majority to pass routine legislation instead of 1/2. I don't remember the exact phrasing now, but it took a few tries to get a decent answer which said it couldn't tell because a lot of the bills passed by voice votes. It listed the legislation and I verified a couple. That's about as far as I trust AIs.
I think AIs can be reasonably trusted, IFFFF you put a lot of iterative work into rephrasing the question and checking the results to revise the question. The success stories which I trust seem to have spent hours giving it more and more detailed instructions. The problem is that most people don't have the patience for what all the AI brands tout as near instantaneous.
^this^ my experience as well (outside of very cut and dried type statistics)
One must engage an AI bot always on the lookout for it going astray. If you go in without knowing anything about the subjec you're asking about you had better be armed with a good bit of cautious skepticism
So more testing needed?
Counterpoint: Much of radiology paperwork is ICD-10 codes and limb/organ locations and we already employ less advanced AI/machine learning to replace the GED-level transcriptionists used to make up for many doctor's general lack of 3rd grade writing skills.
Medicine has been using and accumulating brute-force "You see, disease detection has a preset limit. Knowing its weakness, I sent wave after wave of my own men at it until they reached the detection limit and shut it down." for decades.
That depends a lot on what AI you're talking about. If you're talking Generative AI, you're absolutely right - it's close to worthless because it's designed (though not intended) to generate hallucinations.
Detective AI systems (like that described above to scan x-rays) have a much higher success rate.
Want to be a libertarian thought expert. Learn the words might, maybe, could and pontificate. Never review your thoughts on the future. Write for Reason.
Reluctantly and strategically.
Most of what primary care physicians do could be replaced by an app and more prescription medications becoming OTC.
The northwestern credential beings (they are not doctors or people) then taught the Ai to unionize, now they pay the ai to do nothing.
Decent point that Northwestern University and Northwestern Medicine make more money off of grants and tax-free land ownership (in downtown Chicago) than they do off of actual medicine and have for years.
Per the current "They're a trust fund that happens to hand out diplomas." thing with Universities in general; they're a REIT that hands out medical degrees.
Two things are 90% of the problem here. The first is government-mandated bloat, which destroys everything it touches. If every person you interact with didn't have to fill out the same form about your ancestry and how many guns are in your house 29 times, they might be able to, you know, provide healthcare. Not to mention all the redundant paperwork processing behind the scenes.
But the biggest cost reductuon in our current system would come from pricing transparency. Particularly if it l were accompanied by a shift to HSA's. If I don't know what a procedure is going to cost and I don't care as long as "insurance is the one paying it", I'm going to end up with a $450 ankle brace that Amazon sells for $29.99.
Funny how we all see the problem here but neither party will fix it. Granted one party created the current broken system, but the other spent half a dozen election cycles running purely on repealing all this bullshit. Yet somehow they just never get around to it. The Evil party and the Useless party. Fuck em all.
There is already a pricing transparency law. Medical providers are ignoring it.
Cite, with links?
Are they required to disclose that they won't give you a price?
The purpose of your primary care physician is to try and guilt you into taking more vaccines. Can AI make you feel dirty for being unclean?
If what you say is true, increased vaccination rates is a strong argument for not using AI.
Because it's good for profits?
"might" being the operative word.
If AI suggests new relationships that that the physician hadn't thought of, that's great. OTOH, it might simply allow the physician to log every word you've said and then produce an unconsidered, thoughtless summary report. My bet is on the latter.
Fortunately, AI tools can make doctors more efficient.
That's not the real value. The real value is for patients to be able to understand their condition better, to track symptoms, and to be able to ask (and have answered) the more complicated questions that never get asked when the patient is the most ignorant person in the room. That's what will improvement treatment outcomes and make them personal.
That use case likely won't happen in the US because of the way our system is structured. The question should be about effectiveness not efficiency.
Not if the AMA has anything to say about it. And they do.
Don't forget that hospitals are required to treat all. They are usually not a charity. I tell my dad when he complains about his high hospital bills that he is paying for all those without coverage who are treated their. He says no, it is the Republicans fault and will only get angry if told otherwise.
I'm astonished by the wide variety of topics Jack is expert on: beef prices, medical innovation, foreign affairs, criminal justice reform. What a dazzling intellect he must have to know so much about so many different things in his mid 20s.
AI assistance means faster diagnosis of potentially life-threatening conditions.
We have literally spent a quarter of a century telling people NOT to google their symptoms.
*enters symptoms into webMD*
*List of possibilities includes kidney cancer.*
*runs to the doctor*
"I think I might have kidney cancer!"
"Why's that?"
"I googled my symptoms."
".... sigh."