- A-Doc's AI, designed to save lives in emergency rooms, faced unexpected user adoption challenges.
- Initial assumptions about algorithm errors or technical glitches were disproven by deep dives into user behavior.
- Notification fatigue and a lack of relevant context were identified as the primary barriers to effective AI utilization.
- Integrating with hospital scheduling systems and understanding dynamic user workflows transformed the AI's impact.
A-Doc, a pioneering healthcare AI company, faced a perplexing problem: their highly accurate AI system, designed to flag critical medical findings, wasn't being effectively utilized by radiologists. Despite flawless algorithm performance, doctors were not engaging with the life-saving notifications. This deep dive into their discovery process reveals a universal truth for product managers: superior AI results are meaningless without an excellent user experience tailored to real-world contexts.
A-Doc's mission is critical: to leverage AI to improve patient outcomes in emergency settings. Their technology addresses key challenges like prioritizing urgent cases, accurately diagnosing complex conditions from vast imaging data, and ensuring swift communication to treatment teams. However, despite the advanced capabilities of their computer vision algorithms, the company encountered a significant hurdle: radiologists, overwhelmed by increasing patient loads, were not consistently using the AI tools designed to assist them. The core question became: how do we empower them to do their job better if they're not engaging with our solutions?
The specific issue surfaced with AI-generated notifications for critical findings. On paper, a doctor receiving an alert about a potential stroke or acute condition should immediately act. Yet, A-Doc observed low engagement. The intuitive first reaction was to suspect the algorithm's accuracy, perhaps a high rate of false positives, or technical issues with the notification delivery. However, rigorous internal testing and validation by A-Doc's team of radiologists confirmed the AI's precision. The problem lay elsewhere: the algorithm was perfect, but the user experience was not.
The breakthrough came from understanding
“The core is not the algorithm at all. The algorithm, they have no problem with it, they didn't say it's inaccurate, I don't believe it. The core is, notify me when it's relevant to me, at the right time for me, with the information I want to see.”
- Adi Kaiser, Senior Product Manager
