This job has become the ultimate case study for why AI won’t replace human workers

By Lisa Eadicicco, CNN

(CNN) — Want to understand how artificial intelligence could change your job? Look to radiology as a clue.

Radiology has become a recent talking point in the AI race. It was mentioned multiple times last month by tech executives at the World Economic Forum in Davos as well as in a White House whitepaper about AI and the economy.

Radiology is far from being the only occupation impacted by AI, which is gradually integrating into the work of software engineers, teachers and even plumbers, among many others. If widely adopted, Goldman Sachs estimates that advancements related to AI could displace 6 to 7% of the US workforce, although the technology is expected to create new jobs too.

But the radiology field has become a case study for how AI could enhance, and not replace, jobs. The type of work in radiology is also ideal for AI assistance, said Dr. Po-Hao Chen, a doctor specializing in diagnostic radiology at the Cleveland Clinic.

Radiology has plenty of available data for AI research and applications, which need copious amounts of data for training. AI can parse through troves of data much more quickly than human workers can, and it is already helping to speed up certain processes in radiology — for example, figuring out which scans need immediate attention.

But human physicians are still required to do the bulk of the work – like making diagnoses, physically examining patients and writing reports. And radiology jobs are projected to grow faster than roles in other areas as the field continues to embrace the tech.

“(AI) is not only not replacing those workers, but it’s actually increasing the amount of work they can do and increasing demand for their services,” said Jack Karsten, a research fellow at Georgetown’s Center for Security and Emerging Technology. “That’s sort of a bright future that the tech industry can point to as far as this is AI doing good in the economy.”

How AI helps a job without replacing it

AI is very good at analyzing images and spotting patterns in data, both critical to radiology. And the field has been digitized for years, meaning there is an abundance of data, according to Chen.

“There are smaller use cases that are analogue still, but in the US for the most part, every X-ray, every CT (scan), every MRI, can be available as zeros and ones,” Chen said.

Today, radiologists are using AI to help figure out which scans to prioritize, enhance image quality and assist with summarizing reports, according to Dr. Chen and two other radiology experts who spoke with CNN.

“It’s something that doesn’t replace anyone, that just makes our job more efficient and more meaningful,” said Dr. Shadpour Demehri, who works in interventional radiology at Johns Hopkins Medicine.

René Vidal, a professor in engineering and radiology at the University of Pennsylvania’s Penn Engineering department, views AI as particularly useful for capturing high-quality MRI scans with fewer measurements. That speeds up the process and allows more patients to be seen in the same amount of time.

Other applications are being explored in research, such as using AI to measure the volume of a tumor or automatically populate reports, although they’re likely still far out, said Vidal.

Jobs that were predicted to vanish, but didn’t

AI tools must be approved by the US Food and Drug Administration for medical use, which could take around eight years considering the development process and clinical testing, Vidal said. But those approvals are certainly happening: Of the 1,357 AI-enabled medical devices currently with FDA approval, 1,041 are for radiology.

At the same time, radiology jobs seem to be growing. The Bureau of Labor Statistics projects employment in radiology will grow 5 percent from 2024 to 2034, which is higher than the average of 3% across all occupations. Data from Indeed provided to CNN also indicates there were more radiology jobs in 2025 compared to five years ago.

Demand for imaging during the medical diagnosis process, along with an increased aging population, is likely driving the need for more radiology services, say the radiology experts who spoke with CNN.

But that wasn’t always the thinking. Nobel Prize-winning computer scientist Geoffrey Hinton, also referred to as the godfather of AI, said in 2016 that “people should stop training radiologists now” because deep learning – a subset of AI that models how the human brain learns – would handle the job better in five to 10 years.

Hinton said in an email to the New York Times last year that he spoke too broadly in those 2016 comments.

Demehri recalls there being a sense of anxiety in the radiology field about AI replacing human roles around the 2015 and 2016 timeframe. Now, the technology is seen as a “second set of eyes,” he said.

Pitfalls of overreliance

Still, there are risks around bias and potential overreliance on AI, according to Chen. Unlike human radiologists, for example, AI can accurately predict a person’s race based on an X-ray, according to a 2022 MIT study, raising concerns about bias in diagnoses.

Chen says he also worries about the temptation to make staffing decisions – such as replacing a doctor with a nurse or a subspecialist radiologist with a primary care doctor – if AI becomes advanced enough. That might work in some cases, but not for the majority of conditions that radiology is used for, like detecting cancer or deadly infections.

“We have to understand that a lot of the performance of (the) algorithm comes from the fact that the automation output is reviewed by an expert,” he said. “And together, this collaboration, if you will, between the machine and the expert is what makes the improvement real.”

The-CNN-Wire
™ & © 2026 Cable News Network, Inc., a Warner Bros. Discovery Company. All rights reserved.