A “Companion at Your Shoulder”: Diagnostic Support Inside the Point-of-Care Workflow
Radiology has no shortage of information, instead the challenge is accessing the right information at the right moment without disrupting workflow.
That was the central theme of a recent webinar discussing the RSNA Ventures and Rad AI partnership. The initiative focuses on embedding curated RSNA journal content directly into Rad AI Reporting, allowing radiologists to access in context reference material within the reporting environment at the point of interpretation.
As Demetri Giannikopoulos, Chief Innovation Officer at Rad AI, described it, the goal is to take “robust journal articles and scientific literature, make it actionable, and then … integrate it directly into the radiologist’s workflow.”
Below are a few early insights from the panelists on the potential benefits of in-workflow trusted insights for quality, efficiency and consistency.
1) Embedded Reference Access Versus External Lookup
Radiologists routinely consult external resources during interpretation. Melissa Davis, MD, MBA, Vice Chair of Medical Informatics at Yale University, stated: “I still look up information online all the time … multiple, multiple times a day when we come across something that we want to reference.”
The issue isn’t access to information but workflow interruption and source reliability.
Adam E. Flanders, MD, RSNA Board Member and Professor of Radiology at Thomas Jefferson University, explained that radiologists “don’t have enough time to actually do formal lookups,” and when they do, “how do we know the veracity of that information? … Do we trust it?”
The intent of the partnership is to provide “relevant RSNA content … just enough to get you to move on to your next step.” Rather than presenting full articles, the system is designed to surface targeted content – classification systems, differential considerations and referenced material – based on the context of the report.
Dr. Davis emphasized the importance of trust in this setting: “Having a tool like this directly embedded within the platform that also is trustworthy is extremely important.”
2) Cognitive Load and Workflow Impact
The webinar addressed cognitive load as a practical concern in modern radiology practice.
Dr. Davis noted that it’s “quite difficult to directly measure cognitive load because it really is about how your radiologists are feeling on a certain day.” Yale University evaluates impact through surveys and user feedback, including a straightforward question: If the tool were removed, would radiologists want it back?
Regarding generative AI impressions, like Rad AI Impressions, she commented that for complex reports, “it just feels better … when I can review a finding and edit it, rather than having to regurgitate what I just said.” The implication is that reducing repetition and minimizing workflow disruption can improve day-to-day experience, even if quantitative measurement is challenging.
3) Operational Considerations in Community Practice
Joseph Guiffrida, Chief Operations Officer at ARA Health Specialists, discussed the operational implications of embedded support.
He described prior efforts to centralize reference material in shared folders or internal systems. Usage depended heavily on available time. “If their hair’s not on fire,” radiologists might navigate to those resources, but under volume pressure, they often did not.
Embedding support within the reporting and voice recognition workflow changes accessibility. According to Guiffrida, the expected benefits include “fewer addendums, fewer callbacks…more standardization of our reporting,” with the goal that referring clinicians receive “the same report … every time, regardless of the radiologist that read it.”
He also noted interest in measuring impact through addendum rates, incoming calls and standardization metrics.
4) Why This Moment Is Different
Dr. Flanders, who has experienced prior technology transitions in radiology, commented on the timing of this initiative: “If our volumes were not up 300-plus percent compared to 10 years ago, people might not care as much.”
Rising volumes, staffing pressures and expanded coverage expectations have changed the tolerance for inefficiency. Radiologists continue to seek high-quality performance while managing increasing throughput demands.
Dr. Davis added that expectations around usability are evolving. Radiologists frequently compare clinical systems to consumer technology, asking why healthcare platforms don’t offer similar functionality.
5) The Companion Concept and Defined Boundaries
A central metaphor discussed during the webinar was that of a companion. Dr. Flanders described the embedding of RSNA information into Rad AI Reporting as “sitting on your shoulder saying, don’t worry, I got you covered here,” anticipating needs such as classification systems or differential considerations without requiring separate searches.
Importantly, he clarified the intended scope. The tool “isn’t designed to draft a report or generate a diagnosis.” Instead, it’s intended “to give the radiologist the optimal information they need … to broaden a differential or provide a classification system.”
The system functions as embedded reference support rather than autonomous clinical decision-making.
Trusted Intelligence That Fits the Radiologist Workflow
Radiology doesn’t need more information — it needs the right information, delivered at the right moment without disrupting interpretation. The RSNA Ventures and Rad AI partnership reflects that shift.
By embedding curated, trusted RSNA content directly into Rad AI Reporting, diagnostic support can move from something radiologists have to seek out to something that supports them in real time.
To see how the diagnostic support companion functions in practice — including a live demonstration and deeper discussion of implementation strategy — watch the full webinar recording.
If you’re evaluating workflow tools for your practice and want to explore how Rad AI Reporting can support standardization, efficiency and trusted reference access, we welcome the conversation.
