Half of Our Expenses Go Into R&D. Your Lab Gets the Result

Half of our expenses go into R&D. Your lab gets the result.

At MovixTech, we invest 50% of our expenses in R&D. 50 cents of every dollar we spend.

Lab owners benefit from it.

Accuracy across four generations

Here is the accuracy of 4 generations of our Scan Quality Control AI Employee:

• V1: 77%
• V2: 68% (too many false positive results, the AI model was too strict)
• V3: 87%
• V4 (released on June 22, 2026): 96%

The accuracy of a dental technician doing QC is 80-85%. As you can see, V3 was already more accurate than a dental technician, but we continued to invest in R&D.

If you hire an AI Employee, you increase the accuracy and consistency of QC in the lab.

Why we built our own AI models

We built our own AI models: think ChatGPT for dental scans. General-purpose LLMs have low accuracy on dental scans. Fine for a demo. If you need high accuracy and consistency, you need your own models.

There is no foundational model for 3D dental geometry you can prompt your way into. We’ve built two foundational AI models. These are the brain of our AI.

Why we label data in-house

Another bottleneck is labeled data from a domain where almost nobody has both the clinical judgment and the manufacturing context. We teach our AI models what is good and what is bad (a good crown prep, a bad crown prep, and what is in the middle: what is fixable).

We did the data labeling internally. Strict requirements for the team: our labeling technicians have 6+ years in digital dentistry on average. People who spend 8 hours a day working with digital models, not the ones who touch digital dentistry from time to time.

Garbage in, garbage out. We take that seriously. The quality of the dataset shows up directly in the work of our AI Employee.

The picture below is that dataset. Our AI researcher put it together.

We tried to outsource labeling. The quality was not there. So we decided to spend the money and the time in-house.

What that money bought over the last 12 months

• About 20 releases. Every other week, across both AI Employees.
• We retrained our foundational models twice (January 2026 and July 2026) and keep improving them.

What it looks like in the lab

• Two AI Employees live, the first in the world.
• AI Employees do the job. Not a feature, not a bot. They get the job done.

500 consecutive production cases, reviewed case by case by senior dental technicians with 6+ years in QC and scan repair:

• 97% of defect-free scans cleared. No false retakes.
• False positives ~3% overall, under 1.7% per defect type.
• ~95% agreement with expert human review.

We are confident that spending 50% of our expenses on R&D is what lets us build an AI Employee that does the work end to end. Not just another piece of AI software.

Join the webinar

And when AI Employees do the work end to end, it shows up in the numbers. On October 29, in a closed-door MovixTech webinar, Dr. Ellen Polsky, founder and CEO of DashCare Group, a 34-location DSO, will walk dental lab owners through her own rollout and what it saved.