To keep ourselves competitive in an industry where speed and efficiency lead to better medical decisions, we created a tool that would lean into AI and deep learning tools—something that is hard to achieve and takes years to get FDA-approved. This path also made sense because we could build on our successful in-market products, targeting customers who didn't renew their subscription or were unlikely to adopt in the first place. We worked closely with our deep learning and imaging teams to balance what was achievable but also simple to pass regulatory.

Doctors will always communicate their desire for an easier and more efficient solution to make clinical decisions faster. However, the American healthcare system is mired in bureaucracy. The challenge, though, for experienced doctors is that their habits, behaviors, and tools are so well-ingrained in their current routine and infrastructure. But for students and doctors new to ultrasound, these advanced exams could jump start their training, leading to longer-term inclusion of ultrasound in their clinical workflow.

Over three quarters, we focused most of our iterations on content development and AI tools (autocapture, quality indicator, and image overlay). We made many adjustments to our schedule and deadline to account for the difficulties with our AI tool accuracy. We learned rather quickly that we weren't nailing the efficacy of the autocapture or image overlay features–key features which delivered the core value of this new product.

Finding the right balance of length and accuracy within the content framework was hard to get right as were the AI functionalities. Mobile phone tools can be tough experiences because of the screen size—but this platform made sense for active users whose shared goal is to capture scans across patients in different locations. In terms of content, scanners are splitting their focus between the instructions, the probe, and connecting with the patient. And the AI features had behaviors that were especially hard to stabilize through multiple rounds of research.

We learned so much about the limitation of our advanced AI tools through this project, having to pause on autocapture and the quality indicator and outright remove image overlay. Eventually, with the reprioritization of products that could be taken to market much faster, Advanced Exams were put on the shelf to be revisited at a later time when the business could take bigger risks on projects with longer release timelines or lower-confidence commercial wins.