Photo: Getty Images
Artificial intelligence (AI) has enhanced diagnostic accuracy and improved triage in radiology. But far fewer tools address the daily inefficiencies that consume a radiologist’s time.
“In some ways, detection and triage algorithms can slow radiology down, because we need to track down all of the detections and determine which ones are false and which ones are true,” says Curtis P. Langlotz, MD, PhD, the director of the Stanford Center for AI in Medicine and Imaging (AIMI) and past president of the Radiological Society of North America.
Stepping into the void are AI tools that automate various parts of reporting, a previously manual process that consumed between 75% and 80% of a radiologist’s day, according to Dr. Jeff Chang, the youngest U.S. radiologist in history and co-founder of Rad AI.
Document Time Savings with AI
Reporting tools, Dr. Langlotz says, streamline workflows for radiologists, much like how ambient scribes accelerate transcription during clinic visits. “The doctor at the end of the visit doesn’t have to start with a blank page,” he says. “They have a draft that AI has created, and that saves quite a bit of time and cognitive effort.”
Rad AI pioneered generative AI reporting features with Impressions, a product that automatically generates the radiologist’s impressions from dictations using the doctor’s own language. “On average, it saves about an hour per nine-hour shift and reduces the number of words radiologists would say by about a third,” Dr. Chang says.
Rad AI’s newest product, a full reporting solution, implements multiple generative AI features so radiologists can draft a full report from a few spoken words, auto-populate findings from prior studies, and auto-summarize earlier reports automatically. It can also pull measurements from technologist worksheets using optical character recognition, reducing manual data entry.
“If we can save radiologists an hour a day with [Rad AI] reporting, we can help improve productivity by up to 25% or more,” Dr. Chang says. “That’s still not enough to cover the [radiologist] shortage, but it can help to reduce the burden on radiologists.”
Dr. Langlotz and his team at the AIMI Center have researched the estimated impact of reporting and other AI tools on the radiology workforce. “What we show is that they do, in fact, help us cope with rising demand, so if we become more efficient, there will be a lower need for radiologists’ time in the next many years,” he says. “This is one more technologic innovation that’s going to help us work more efficiently and absorb some of the increasing volumes we’re seeing.”
Beyond Reporting Tools
While AI-enabled reporting apps show the greatest potential efficiency gains, Dr. Langlotz views other emerging tools as workflow enablers, too. Newer agentic AI tools show promise in streamlining clerical tasks, such as prefilling measurements or extracting information from ultrasound reports and adding it to radiology reports automatically.
Unified radiology ecosystems, meanwhile, can reduce context switching and alleviate the reality that, as new tools emerge, “radiologists become the primary mode of integrating all of those things on the desktop,” Dr. Langlotz says.
Dr. Langlotz is also watching the development of autonomous AI tools, while cautioning that human radiologists should always remain in the loop.
“A lot of work remains on what is the optimal way for radiologists and AI to work together, but I think the combination of human and machine is better than either one alone,” he says. “The machine is going to be better at some of the routine tasks like measurements, and humans will be better at thinking about the pathophysiology, interpretation, thinking about what the best therapy might be, and communicating that back to clinical teams.”
The communication piece, Dr. Chang adds, is essential. “Part of our goal is to return radiologists to being the ‘doctor’s doctor,’” he says, “to give the radiologist more time to review and analyze images and more time to discuss directly with the ordering physician and figure out exactly what they need versus doing nonstop dictation.”
July 10, 2026 
