One morning in April, radiologist Andrew Moriarity was listening to a podcast episode about self-driving cars when the conversation began to sound strangely familiar.
“In the early days, self-driving cars actually needed twice as many people as a normal car to drive,” Moriarity said. “So I think we’ll get there, but we’re much closer to 2010 for healthcare AI than we are to 2026.”
Moriarity’s analogy speaks to a hot-button issue for radiologists across the country: the integration of AI tools into their clinical workflows.
As the U.S. population is aging and cancer rates are rising, the growing demand for X-rays, MRIs and other costly imaging tests can no longer be met by the radiology workforce. Artificial intelligence has been touted as a way to boost productivity and reduce labor shortages, but new algorithms for medical imaging require extensive oversight and validation. As things currently stand, AI isn’t delivering fast enough to ease the high costs, heavy workloads and capacity bottlenecks in the field.
“The demands on imaging have changed so dramatically in the last 10 years,” said radiology consultant Tobias Gilk. “We’re at the point now where baby boomers have begun crossing the 65-year-old threshold in significant numbers.”
According to the National Institutes of Health, seniors utilize medical imaging at a rate two or more times higher than the rest of the population. But aging isn’t the only factor behind the growing workload of radiologists — liability concerns are also contributing to record-high imaging volumes. According to a 2024 survey, more than 38% of radiologists have faced a medical malpractice suit at least once in their career, a rate higher than most other specialities.
Yet most radiologists do not often order images on behalf of patients. Rather, they assume the downstream burden of scans ordered by emergency and primary care doctors seeking to reduce their own legal risk.
“Imaging can be kind of a diagnostic shortcut,” said Eric Christensen, research director at the Harvey L. Neiman Health Policy Institute. “If you’re an E.R. doctor and you don’t really have time to do a full physical evaluation, it’s easier to just send the patient through the scanner.”
With legal and demographic pressures beyond their control, hospital systems and radiology practices are increasingly turning to AI tools. In 2025, around 30% of U.S. radiologists used AI in their routine clinical workflows. Among them was Moriarity, diagnostic radiologist and president of Advanced Radiology Services in Grand Rapids, Mich.
Though Moriarity describes himself as an AI optimist, he says that current algorithms create more work for the practitioners who must manually review and verify their outputs.
“It doesn’t make my job easier — it adds complexity, it adds work, and it adds more to what we call the cognitive or mental load,” said Moriarity. “I have to make more decisions every single day because of the AI.”
Moriarity cited hallucinations and false positives as major issues with AI — and he’s far from alone. A 2024 survey from the American College of Radiology found that while 70% of radiologists had positive perceptions of AI tools, only 46% reported that the overall accuracy of AI met their expectations.
Faster diagnoses and growing workloads
The radiology AI market was worth $14.6 billion in 2025, and this number is expected to increase more than ten-fold by 2033. One of the companies fueling this growth is Qure.ai, which develops AI tools to assist radiologists in interpreting medical images. Founded by data scientist Prashant Warier in 2016, Qure.ai now boasts hundreds of employees, global offices and 26 FDA approvals for its algorithms, which have been implemented in more than 2,100 American hospitals to date.
Qure.ai’s products have shown an over 95% success rate in detecting the small nodules that appear in the earliest stages of lung cancer. As opposed to the productivity of radiologists, early detection remains the company’s ethos.
“Our value proposition is more about early diagnosis than reducing workload,” Warier said. “Our goal is to get the patient through the system faster and diagnose patients who are not getting diagnosed early.”
The number of FDA-approved, AI-enabled radiology products has skyrocketed over the past decade. Only 25 such devices were on the U.S. market in 2016, compared with 253 new approvals in the last year alone.
While the benefits of certain algorithms are undeniable for patients, radiologists themselves are less likely to share in those gains for some time. In fact, some AI tools will increase the volume and pace of work they are expected to handle moving forward.
Shifting expectations and uneven impacts
For now, radiologists are more concerned about AI reshaping their workflows rather than replacing them altogether. As AI raises expectations for radiologists to review more images in less time, its impact will vary across the field.
“There will absolutely be folks whose jobs get much easier, and folks whose jobs stay the same or maybe even get harder,” said Gilk. “And I don’t think anyone today is in a position to crystal ball and give us any solid indication of who those winners and losers are likely to be.”
Newer radiologists are starting to train with AI-integrated workflows, but a knowledge gap remains among those who have long practiced without them. While established radiologists worry their lack of familiarity with AI tools could impact their job security, newcomers are wondering what their careers will look like in the years ahead.
This question is front of mind for Anna Gregg, a recent medical school graduate in Nevada. As she prepares to begin a radiology fellowship next year, she recently completed an elective course on AI applications in the field.
Gregg feels cautiously optimistic about the long-term potential of AI, but she is less convinced about its near-term ability to support practitioners.
“AI is just not where people think it is,” Gregg said. “It’s not going to take over jobs. It’s not going to help radiologists read scans quicker and it’s not going to solve the labor shortage of radiologists.”
