5 Questions Every Radiology Practice Should Ask Before Choosing a PowerScribe® Replacement
Vendor selection is where the PowerScribe® 360 migration gets complicated.
The market has responded to the end-of-life notice the way markets typically respond to a disruption: with a surge of new entrants, expanded feature lists and confident claims about AI capabilities. For radiology leaders trying to evaluate meaningfully under real-time pressure, the noise can make a clear decision harder, not easier.
The panel in our recent webinar — a chief information officer who has already completed this migration at scale, a practicing neuroradiologist and implementation and customer success leaders who have guided organizations through it — converged on a shortlist of questions that cut through the hype. While this isn’t a comprehensive RFP checklist, these five questions surface the real differences between vendors and options.
Can you scale to meet us where we are — and where we're going?
Tom Hasley, CIO of LucidHealth, asked this question, and it's the right one to lead with. A speech engine that performs in a controlled demo environment isn’t the same as one that performs accurately across regional accents, subspecialty terminology and hundreds of concurrent users under mission-critical uptime requirements. Many new entrants, and legacy vendors alike, can build something that works in a bubble. Scaling it is a different challenge entirely — and the organizations that find that out after go-live are the ones that struggle most.
Who’s leading product design, and how involved are radiologists?
Any vendor can claim their platform is built for radiologists. The more useful question is whether radiologists are leading design and development, meaning they’re embedded in the process, not consulted after the fact. Ask for specifics, like how many clinical advisors are involved? At what stage? What has changed in the product as a direct result of radiologist feedback? Rishi Seth, MD, a practicing neuroradiologist and CMIO at Rad AI, said the answers reveal whether clinical input is structural or decorative.
What does support look like in the first 90 days — and what does customer success look like after that?
The gap between go-live support and ongoing customer success is where implementations quietly fail. Understand exactly who your point of contact is, what metrics they monitor proactively and how they engage when something isn't working — before you log a ticket. Joseph Mack, Vice President of Customer Success at Rad AI, said the difference between a dedicated customer success manager who is watching your usage and performance data and a tiered ticketing system is significant. It shows up in the adoption curve.
What infrastructure is required, and will our IT team need to manage additional systems?
Paul Lipton, Lead Product Manager at Rad AI, said cloud-native platforms should reduce IT burden, not redistribute it. Before signing, understand the full scope: licensing requirements, hardware requirements, integration dependencies, ongoing maintenance obligations and what happens when something breaks. The total cost of ownership conversation needs to happen before contract, not during implementation.
Can I see your AI governance documentation and model cards?
With regulation actively evolving and patients now receiving reports in real time under the 21st Century Cures Act, how a vendor builds, validates and governs their models is a clinical and legal consideration — not just a procurement checkbox. Demetri Giannikopoulos, Chief Innovation Officer at Rad AI, said vendors who can answer this question readily and transparently are worth paying attention to.
The full panel conversation — including the context behind each of these questions and the Q&A that followed — is available via the on-demand webinar.
