Three CMIOs, Three Perspectives: How Radiology Leaders See the Future of AI

If you want to understand where AI in radiology is truly headed, ask the people shaping it from within. At Rad AI, our three Chief Medical Information Officers sit at the intersection of clinical practice and technological innovation translating the realities of radiology into smarter, safer, and more scalable AI solutions.

In this candid roundtable, Dr. Elizabeth Bergey, Dr. Rishi Seth, and Dr. Andrew Del Gaizo tackle the same three questions: how AI can make radiologists’ value more visible, when to build versus buy, and what success will look like five years from now. Their perspectives reveal not just where AI is going, but how it can strengthen the connection between radiologists and the patients they serve.

Much of what radiologists do is invisible to patients and even colleagues. How could AI either amplify or further obscure the value radiologists bring?

Dr. Elizabeth Bergey: Radiologists are often considered the doctor’s doctor because we’ve traditionally served as consultants to more clinically focused physicians. But as radiologist shortages have become widespread and intensified, many of those valuable doctor-to-doctor interactions have been minimized, if not eliminated.

AI has the potential to amplify the value radiologists bring to patient care by freeing up time currently consumed by clerical duties. That extra time can lessen the strain radiologists feel and allow for more meaningful consultative interactions with referring physicians.

It also helps make radiologists more visible to patients. With patient portals now ubiquitous, AI can help translate radiology reports into lay terms. When patients read and understand their reports, they discover a previously unrecognized member of their care team, the radiologist, who is no longer obscured by the darkness of the reading room.

Dr. Rishi Seth: AI can amplify radiologists’ value by removing clerical friction—surfacing priors, reviewing documentation, structuring findings, and drafting clear impressions—so radiologists spend more time on image interpretation and clinical judgment.

But AI can also obscure that value when it blurs the line between human expertise and automated output. Over-automated reports or templates risk hiding the radiologist’s reasoning, the very insight that defines our contribution to patient care.

Dr. Andrew Del Gaizo: Integrating AI into the radiologist’s workflow has the potential to make the invisible visible. When used effectively to route the right study to the right reader, surface priors and relevant labs, draft a coherent impression, or automatically closes the loop on follow-ups, patients and clinicians feel the difference: faster answers, clearer recommendations, and fewer dropped balls.

The risk comes when AI adds clicks, slows us down, or makes radiologists second-guess themselves. My litmus test is simple: does this tool give me time back to think and talk to clinicians/patients, or does it make me feel like a cog in the machine? If it’s the latter, I don't use it as a radiologist, and I don’t authorize deploying it as CMIO at Rad AI.

How do you weigh the decision to buy a vendor AI solution versus building something with your own team or an academic partner? Are there clear signals that favor one approach over another?

Dr. Elizabeth Bergey: I actually did this! Our complex group – multiple hospitals, PACS, RIS, and EMRs – was in the market for a workflow orchestrator. We shopped around at RSNA, but every option we found was essentially a highly filtered worklist. The "buy" option simply didn’t meet our needs, so we did a “make analysis” to estimate the cost of building what we needed.

We needed a dynamic solution that could adapt to the craziness of everyday radiology: knowing which emergency departments were busy, which doctors were unavailable, which cases needed to be read first, etc. After running the numbers, we realized building our own would deliver ROI under two years, and it did.

The benefits went beyond efficiency. Our custom solution also gave us deep analytics on performance and margins across sites and procedures, insights we could never have gained from an off-the-shelf product.

Dr. Rishi Seth: Buy when the solution is well-integrated and delivers quick ROI. Build when the capability is unique or strategically differentiating, like customized reporting or domain-specific speech-to-text. 

The decision often hinges on data access, iteration speed, and long-term ownership of the workflow. In a practice setting, it’s equally important to consider who will maintain and support the solution over time.

Dr. Andrew Del Gaizo: We’re all stretched thin these days. Therefore, I’d say buy when the problem is common, regulated, and requires enterprise-grade uptime, integrations, support, and liability frameworks. 

Build when the problem is truly differentiating for your practice, when the data is uniquely yours, and when you can sustain model operations, validation, and security.

Signals that favor buying include proven results from other practices, clear ROI, and a vendor that truly cares about their customers and is committed to governance and success metrics. 

Signals that favor building include niche workflows that vendors cannot financially prioritize, or research questions where the speed of iteration is more important than feature completeness, with the option to bring the solution into production if it proves successful.

Let’s fast forward five years. What do you hope radiologists will say AI finally got right – something that makes them wonder how they ever worked without it?

Dr. Elizabeth Bergey: I hope AI reduces the clerical burden radiologists face every day. It should premeasure lesions, compare measurements to prior studies, and generate tables or graphs that show changes over time. It should draft reports within each radiologist’s preferred template and route normal exams to a preliminary normal queue.

When AI can take care of those repetitive tasks, radiologists will finally be free to focus on what truly matters: clinical interpretation and patient care.

Dr. Rishi Seth: As someone who is deeply focused on user experience, I hope they will say AI finally disappeared into the workflow. AI should seamlessly exist within our PACS and reporting applications. The tools should feel like intuitive extensions of the radiologist rather than separate systems. Radiologists will wonder how they ever worked without that ambient, intelligent assistant.

Dr. Andrew Del Gaizo: My hope is that AI exponentially delivers on its efficiency promise, giving time back to the doctor to look at the image rather than perform clerical duties. 

Further, I hope AI brings clarity to the decisions we make (less hedging) and peace of mind that downstream care will be appropriate (no patient lost to follow-up). I think a big step toward this will require explainable AI, where AI flags its uncertainty and knows when to step aside.

Finally, I see a world where the joy of providing patient care returns. If we spend less time hunting for data and performing administrative tasks, there is tremendous potential to reduce the burnout symptoms many doctors experience. 

Personally, this is why I leaned into the CMIO/entrepreneurial path: to create tools that give radiologists their time and their voice back. When we do that, the altruism that drew many of us to medicine can shine through.