In acute ischemic stroke, every minute without treatment results in the death of approximately 1.9 million neurons. The phrase "time is brain" is not a metaphor — it is a quantified biological fact. CT angiography can identify large vessel occlusion (LVO) strokes in minutes, but getting the right specialist to see that CT scan and mobilize a response has historically taken much longer. Viz.ai looked at that gap and built a product entirely around closing it. The result is not better radiology AI — it is a communication platform with AI as the trigger.
What Viz.ai Actually Built
When a CT angiography scan is completed for a suspected stroke patient, Viz.ai's algorithm analyzes the images in real time, identifies LVO patterns, and immediately sends a push notification — including the actual CT images — directly to the on-call neurovascular specialist's phone. Before Viz.ai, the workflow looked like this: technologist completes scan, uploads to PACS, radiology reads it (30-60 minutes later at night), calls the neurologist, who then mobilizes the interventional team. Viz.ai collapses three handoffs into one automated notification. The clinical validation showed a median 26-minute reduction in time to treatment decision.
That 26-minute number is the entire business. Translated into mortality and functional outcomes at scale, it justifies the hospital's subscription cost, the physician champion's advocacy, and the regulatory burden of going through De Novo 510(k) clearance. Viz built their go-to-market story entirely around a single number that every stroke neurologist immediately understood and cared about.
The Product Insight That Others Missed
Multiple companies were working on stroke AI in 2017-2018. Most of them were trying to build a better AI radiologist — a model that could diagnose strokes more accurately than human radiologists. Viz.ai's insight was that accuracy was not the bottleneck. Radiologists were already highly accurate at reading CT angiograms. The bottleneck was notification latency and care team coordination. They built a product that solved the actual bottleneck — workflow communication — rather than a technically impressive solution to a non-bottleneck problem.
This is one of the clearest examples I know of what it means to do clinical workflow discovery before building. If you ask a radiologist what their problem is, they might say "I have too many scans to read." But if you map the end-to-end stroke treatment workflow with a stopwatch, you find that radiology reading time is not where the delays are. They are in handoffs and notifications. Viz interviewed the stroke neurologists, the interventional teams, and the ED physicians. They found the bottleneck by watching the workflow, not by asking about the radiology AI problem.
The Business Model and Regulatory Strategy
Viz charges hospitals a per-facility annual subscription, typically in the range of $75-150K depending on volume and configuration. The ROI case is simple: if a hospital treats 200 LVO strokes per year and each 26-minute time savings reduces average length of stay by 0.5 days, the subscription pays for itself in avoided costs before counting the outcome improvements and liability reduction. The subscription model also aligns incentives correctly: Viz's revenue does not scale with individual scans, so there is no pressure to flag more positives. The business only works if the product is used and trusted.
The De Novo 510(k) regulatory pathway was a strategic choice. Viz could not find a predicate device — there was no prior cleared device doing what they were doing — so they went De Novo, which establishes a new device category. This was slower and more expensive than a traditional 510(k), but it created a classification order that subsequent competitors had to follow, giving Viz first-mover advantage in how the FDA frames the product category. That regulatory investment was a competitive moat, not just a compliance checkbox.
