Each year, more than 40,000 professionals and over 600 exhibitors gather at the Radiological Society of North America (RSNA) annual meeting. Housed in the largest event space in the USA, the Chicago-based event is by far the the most important of its kind, and the place where the future of radiology takes shape.
Once dominated by scanners and imaging hardware, RSNA has since become a mirror of a much larger global shift: a profession standing at the threshold of workflow automation on an unprecedented scale.
Radiology has always been one of medicine’s technological pioneers – from the first digital images to front-end speech recognition. Now it is again leading the way as Large Language Models (LLMs) and intelligent automation begin to transform the everyday workflows radiologists depend on.
The conversation about artificial intelligence in radiology has matured. The topic no longer simply concerns what AI can detect in images, but rather how it can help radiologists confront the greatest challenge they face: overhwhelming demand.
From fragmented steps to continuous intelligence
Globally, the demand for radiology is surging. Hundreds of millions of exams pass across radiologists‘ workstations every year, rising in line with a demographic shift to larger, older populations. Every exam is loaded with data and portfolios of images to study, and the complexity of cases is only increasing. At the same time, the total number of radiologists to do the work has flatlined for the last decade.
According to studies, up to 60% of radiologists are suffering from burnout, 50% of radiologists switched practices in the last year, and over the last 13 years there has been a 50% loss of professional reimbursement.
So at events like RSNA 2025 – taking place this year over the first week of December – the focus will continue to shift from image interpretation to workflow orchestration and automating the small, repetitive steps that consume so much of a radiologist’s day.
LLMs sit at the center of this transformation. They can interpret clinical context, retrieve prior reports, summarise findings, generate standardised text, and anticipate the next logical action, whether that’s suggesting a follow-up or creating a referring letter.
The impact can be immediate: measurable efficiency gains, smoother collaboration, and more time for radiologists to focus on the high-value reasoning that requires their expertise.
“Automation doesn’t replace radiologists. It removes friction from their work,“ says Prof. Dr. Wieland Sommer, radiologist and founder and CEO of Jacobian, a company leading the AI-powered charge towards transformed radiology workflows.
Achieving holistically integrated workflows
Jacobian was founded just this year from the merger of two long-standing pioneers in the field: Pittsburgh-based Fluency for Imaging (FFI) and Munich based Smart Reporting. While it’s a new company with a bold new vision, it is formed from organisations that are already processing more than 80 million exams a year.
FFI, originally known to many clinicians as M*Modal, is a platform that first delivered truly reliable and scalable AI-powered speech recognition into radiology reporting. It helped usher in the end of an era of dictating reports to human transcriptionists and is one of the most widely used reporting solutions in the world. Smart Reporting, founded by Sommer, has further pioneered the application of AI into the formulation of reports themselves, leveraging deep, embedded medical knowledge and guidelines to create standardised, mineable outputs.
In bringing the organisations together, Sommer says the new, larger company can work to empower radiologists with the AI-powered tools they need without causing the harmful disruption that has held back widespread adoption to date.
It’s an approach Sommer and colleagues will be unveiling in detail to the world’s radiology community at RSNA 2025.
“When documentation, reporting, and follow-up run automatically in the background, radiologists can spend more time doing what only humans can – interpreting, teaching, and advising,” he adds.
In many hospitals, radiology workflows still consist of dozens of disconnected actions – dictating, proofreading, formatting, coding, and more. Each hand-off introduces delay and opportunity for error. By embedding automation directly into the reporting environment – powered by LLMs that understand clinical language – many of these manual transitions can now disappear.
Modern systems are evolving from passive tools into intelligent assistants that listen, learn, and act. Measurements, standard phrases, prior findings, even registry data can be filled automatically. The result isn’t just speed, it’s consistency- a key driver of diagnostic quality.
Russ Cardwell, a long-standing leader at FFI and now Senior Vice President at Jacobian, adds: “We start with our deep, client-first commitment and by leveraging a familiar user experience. We’re laser-focused on enabling radiologists to thrive and drive diagnostic excellence for every patient by improving reporting speed and consistency without sacrificing quality.
“We accomplish this by using GenAI tools seamlessly, and by accessing clinical guidelines that our medical content team curates. It’s about innovation without disruption.”
Jacobian’s freshly united teams are using LLMs to automate documentation and data exchange at every stage of the workflow. Instead of acting as isolated systems, reporting environments are becoming connected ecosystems linking voice input, structured templates, analytics, and clinical integration.
Sommer notes that these models go beyond text generation: “The difference between automation and intelligence is understanding,” he says.
“An intelligent system knows not just what was said, but what was meant. When a radiologist recommends a follow-up, the system can recognise that intent, tag it, code it, and route it automatically. That’s how intelligence starts to work in real time, connecting meaning to action.”
The next wave of automation will extend beyond reporting. LLMs can bridge communication between radiology and other specialties by generating referral summaries, tumor-board notes, or registry submissions – tasks that remain largely manual today. Soon, AI-driven orchestration could automatically align radiology workflows with pathology, cardiology, or oncology, creating a shared diagnostic fabric across disciplines.
At RSNA 2025, this evolution is will be visible throughout the giant venue’s halls. Automation is no longer hype; it’s ready to become everyday practice, as Sommer and his Jacobian team intend to demonstrate.
“We’ve reached a stage where technology integrates seamlessly into the workflow,” Sommer says. “It enhances decision making without demanding attention. The workflow itself is becoming intelligent – freeing radiologists to focus on interpretation, not routine.”
