Every dentist who has looked at AI scribes has had the same thought: what happens when it gets something wrong? A wrong tooth number, a medication that was discussed but never prescribed, a finding that was never made. If that ends up in the chart, it is your problem, not the software's. That instinct is correct. It is also exactly why Molaris is built around one rule: the AI never signs the chart.
A draft is not a record
Molaris listens to the visit and produces a structured draft note with eleven sections: chief complaint, medical history, findings and diagnosis, treatment performed, tooth numbers, anesthesia, materials, post-op instructions, prescriptions, next visit, and provider notes. Every one of those sections is a proposal.
Nothing files to Open Dental until a licensed provider reads the note, edits what needs editing, and approves it. That approval step is not optional, and there is no setting that turns it off. Molaris never finalizes a record on its own. Once you approve, the note attaches to today's appointment in Open Dental, the same as a note you typed yourself.
Why auto-finalizing AI is a liability problem
Your chart is a legal record. It is what a malpractice attorney reads, what an insurance auditor pulls, and what a state board reviews when a complaint lands. Every entry carries your signature, and your license sits behind that signature.
Now consider software that writes and finalizes notes automatically. Speech recognition mishears. Language models occasionally state things with confidence that did not happen. In a draft, that is a thirty-second correction. In a finalized record, it is a documented clinical statement you never reviewed, sitting under your name.
The signature exists to mean one thing: a licensed provider read this and stands behind it. Software cannot carry that responsibility. Any tool that quietly removes the review step is not saving you time. It is transferring risk to you.
Signed means locked
The boundary works in the other direction too. Once you sign a note in Molaris, it locks. You cannot silently rewrite it later, and neither can we. If something needs correcting, the correction becomes a separate signed amendment, with the original preserved and a full audit trail showing what changed, when, and who signed it.
That is not bureaucracy for its own sake. A record with disciplined, timestamped amendments is exactly what you want in front of an auditor or an attorney. A record that can be edited invisibly after the fact is worth very little as evidence, even when everything in it is true.
The documentation burden is real, the shortcut is not
Auto-finalizing tools exist because documentation time is genuinely crushing. In medicine, this has been measured carefully. A time and motion study in Annals of Internal Medicine found that physicians spend nearly two hours on EHR and desk work for every hour of direct patient care (Sinsky et al., 2016). A study in Annals of Family Medicine using EHR event logs found primary care physicians spending close to six hours a day in the EHR, including roughly 90 minutes after clinic hours (Arndt et al., 2017). Research in Mayo Clinic Proceedings tied that clerical burden to lower satisfaction and higher rates of burnout (Shanafelt et al., 2016).
Dentistry has not been measured as thoroughly, but the pattern is familiar. Dentists commonly report notes stacking up between patients and getting finished after the last appointment walks out.
The fix is not to remove review. The fix is to make review fast. Reading and correcting a well-structured draft of a visit you just performed takes a fraction of the time that writing it from a blank screen does. You keep the signature. You lose the typing.
BAA first, and your data stays yours
Two more commitments, because trust questions deserve direct answers.
- The BAA comes before any patient data. Molaris requires a signed business associate agreement before real patient data touches the system. Not after a trial period, not once you upgrade. HIPAA requires it, and we treat it as a hard gate. Every office is onboarded personally rather than through self-serve signup, so nobody slips real patient data into an account before the paperwork is done.
- Your data is never used to train AI models. Your recordings and notes exist to document your patients' care. That is the whole job.
Where the AI stops and you begin
Molaris drafts. You decide. Signed notes lock, corrections leave a trail, the BAA comes before the data, and nothing you record trains anyone's model. If AI is going to sit inside a clinical workflow, this is the shape it has to take.
If you want to see what that discipline is worth in hours, run your own numbers at getmolaris.com/roi, or book a 15-minute demo at cal.com/molvo/molaris and watch a note go from recording to signed in Open Dental.