A scan is not good or bad on its own. It is good or bad for what you are making from it. The same scan that is fine for a night guard can be unusable for MARPE.
Most labs pay for scan quality control twice:
Manual scan QC takes roughly 47 hours across the industry, because teams review 100% of cases to find the ~5% that need a doctor rescan and the ~25% that are fixable in house. MOVIX does it in about two minutes, at 92% accuracy on orthodontic cases against 80 to 85% manual.
Our Scan QC AI Employee performs scan quality control for different ortho products. Each product has different quality criteria.
A scan is only good or bad relative to what you are making from it.
The same scan can be perfectly acceptable for one product and unusable for another. A margin that does not matter for a night guard decides whether a crown fits. Missing anatomy at the back of the arch is irrelevant for one case and fatal for the next.
This is why generic scan checking is worthless to a lab. It has no idea what the case is being made into, so it does one of two things. It rejects scans that would have been fine, and sends the doctor an unnecessary rescan request. Or it passes scans that fail later in production, when the remake is expensive.
The AI employee knows what the case is being manufactured into, and checks the scan against the requirements for that product. That qualifier is the entire difference between this and the generic scan QC built into scanners.
The AI Scan Quality Control Employee autonomously:
Think of it as an experienced dental technician reviewing every case the moment it arrives.
This is worth saying out loud, because it sounds like a small feature and it is not.
When a technician needs to explain a scan problem to a doctor, the best they can do is a screenshot, sometimes with an arrow drawn on it, plus a written description. The doctor receives a flat picture of a three-dimensional problem. Often they cannot tell what they are looking at, so they guess at what to rescan, or they call the lab to ask, or they send back a second scan with the same defect.
We send the doctor the model itself, with the problem marked on it. They rotate it, they see it, and they know exactly what to rescan. That is the difference between a rescan request the doctor acts on correctly the first time and one that costs two more phone calls.
It also changes the tone of the relationship. A screenshot with an arrow reads as a complaint. A marked-up 3D model reads as the lab helping, which matters, because the doctor is the customer here — both for you as a lab, and for us as an AI Employee developer.
