On February 11, 2026, we released Version 2 of our Quality Control AI Agent - and it increased real-world accuracy from 75% to 92%.
Here’s the backstory:
When we first tested our QC AI Agent in a controlled environment, it showed up to 98% accuracy across the parameters we benchmarked.
We launched in October. Signed our first customer in December. And then… real life happened.
In production, with cases submitted by real doctors to a working aligner company (raw data, not human in the loop), accuracy dropped to 75% (vs. 98% in testing and 95% in controlled trials).
So we went back to work - no Christmas break - focused on closing the gap between “lab performance” and “real-world performance.”
And we delivered.
Now, with Version 2, our “digital dental technician” achieves 92% accuracy across any defects present in real-life submissions.
This isn’t the end of the journey - our next target is 95%+.
We have an ask for you - could you help us with a quick benchmark?
Do you track QC accuracy in your lab? If yes, would you be willing to share your benchmark (even a range)?
Here’s why we're asking: in real-world production, our “digital dental technician” is currently at 92% accuracy. Our roadmap target is 95%+, but getting those last few points will take months of R&D.
If that extra ~3% doesn’t materially change outcomes for your lab, we’d rather focus that time on the next feature.
Please send us your benchmarks to marina@movixtech.com
