Curious how artificial intelligence might redefine your job? Take a look at radiology as an indicator.
Radiology has recently emerged as a focal point in the AI discussion. It was repeatedly brought up by tech leaders at the World Economic Forum in Davos last month, and was also featured in the White House’s white paper on AI and the economy.
Radiology is hardly the sole profession being affected by AI, which is progressively being integrated into the work of software engineers, educators, and even plumbers, among many others. If widely adopted, advancements tied to AI could displace an estimated 6–7% of the U.S. workforce, according to Goldman Sachs estimates, though the technology is also anticipated to generate new positions.
However, the field of radiology serves as an illustration of how AI can augment rather than substitute human roles. The nature of radiological work is ideally suited for AI assistance, noted Dr. Po-Hao Chen, a diagnostic radiology specialist at the Cleveland Clinic.
Radiology offers a wealth of accessible data for AI training and applications, which require massive datasets. AI can process extensive amounts of data far quicker than people and is already helping to speed up specific radiological procedures—for instance, flagging scans that demand immediate attention.
Yet, human physicians remain responsible for the bulk of the work, such as formulating diagnoses, conducting physical patient examinations, and composing reports. And radiology positions are projected to see faster growth than in many other sectors as the field continues to adopt new technologies.
“(AI) is not only failing to replace these workers, but it’s increasing the volume of their work and boosting demand for their services,” stated Jack Carsten, a research fellow at Georgetown’s Center for Security and Emerging Technology. “This presents a kind of bright future that the technology industry can point toward, given AI’s positive impact on the economy.”
How AI Enhances Work Without Replacement
AI excels at image analysis and pattern detection within data, specialties crucial to radiology. Furthermore, the field has been digitized for many years, ensuring an abundant supply of data, according to Chen.
“There are smaller use cases that are still analog, but in the U.S., virtually every X-ray, CT, and MRI can be processed as ones and zeros,” Chen remarked.
Currently, radiologists are leveraging artificial intelligence to prioritize scans, improve image quality, and assist in summarizing reports, as detailed by Dr. Chen and two other radiology specialists who spoke with CNN.
“This is something that won’t replace anyone but only makes our jobs more efficient and meaningful,” commented Dr. Shapour Demehri, who works in interventional radiology at Johns Hopkins Medicine.
Rene Vidal, a professor of engineering and radiology within the engineering department at the University of Pennsylvania, finds AI particularly useful for obtaining high-quality MRI scans using fewer measurements. This accelerates the procedure, allowing more patients to be examined in the same timeframe.
Other applications are under investigation, such as using AI to measure tumor volume or automatically populate reports, though these are likely still distant from routine operation, Vidal mentioned.
Jobs That Were Predicted to Vanish But Haven’t
AI tools must receive approval from the U.S. Food and Drug Administration for medical use, a process that can take about eight years when factoring in development and clinical trials, noted Vidal. But these approvals are happening: out of 1,357 AI-enabled medical devices currently FDA-approved, 1,041 are designated for radiology.
Meanwhile, radiology job openings appear to be increasing. The Bureau of Labor Statistics forecasts that employment in radiology will grow by 5 percent between 2024 and 2034, which outpaces the 3% average across all occupations. Data from Indeed, shared with CNN, also indicates that there were more radiology job postings in 2025 than at any point in the preceding five years.
The rising demand for imaging in medical diagnosis, alongside a growing aging population, are likely fueling the need for more radiological services, observed the radiology experts interviewed by CNN.
But this wasn’t always the perceived outcome. Geoffrey Hinton, the Turing Award winner also known as the “godfather of AI,” stated back in 2016 that “people should stop training radiologists now,” because deep learning—a subset of AI that models how the human brain learns—would outperform them within five to ten years.
Hinton wrote in a New York Times letter last year that his 2016 remarks were too broadly phrased.
Demehri recalls that in radiology around 2015 and 2016, there was a palpable anxiety about AI supplanting human roles. Now, the technology is viewed as a “second set of eyes,” he said.
Pitfalls of Over-Reliance
Nevertheless, risks persist concerning bias and potential over-dependence 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 biased diagnostic outcomes.
Chen is also apprehensive about the temptation to make staffing decisions—such as replacing a specialist radiologist with a general practitioner, or a physician with a nurse—if AI becomes sufficiently advanced. While this might work for specific scenarios, it wouldn’t be suitable for the majority of conditions where radiology is essential, such as detecting cancer or life-threatening infections.
“We must recognize that a large part of an algorithm’s performance hinges on the automation’s results being validated by an expert,” he asserted. “And it is this collaboration, if you will, between the machine and the expert that actualizes the improvement.”
About the Author
