Why Eva
Built for dentists, by a dentist.
Eva was purpose-built for dentistry rather than adapted from something generic. That difference lives in the details: the tooth numbers you say out loud, the findings you never mention because they were normal, and what happens to a case after the patient says yes.
The problem
You already did the hard part.
By the time a treatment plan is lost, the clinic has already paid for all of it: the marketing that brought the patient in, the chair time, the exam, the radiographs, the diagnosis, and the conversation where you explained it properly. All of the cost is spent, and then the plan quietly goes nowhere.
Not because anyone stopped caring. Because the follow-through lives in memory, sticky notes and someone's messages.
Acceptance is only the first gate, and it is the only part of this anyone measures. What happens after a patient says yes, the visit that was never rebooked and the stage nobody carried, is largely unmeasured, in our own practice as much as anyone else's.
An organised practice already has the pieces. A practice-management system for records. A schedule for future appointments. Lab software for lab items. Good notes so the front desk knows what to say. Every one of them real, and none of them answering the question that matters: where is this case, and what is it waiting on? The thing holding those pieces together is the team's memory.
A future appointment is not case control. Eva is the layer that knows when the treatment sequence breaks.
What we mean by built for dentistry
Six things a general scribe has no reason to know
Moments from a real clinic day, and what Eva does about each. None of them look like hard problems until you have watched one go wrong in your own surgery.
“#14” is not one tooth.
Under Universal it is the upper left first molar. Under FDI it is the upper right first premolar. Same two digits, different teeth, different countries. Eva holds both maps and both ways of saying them aloud: FDI digit by digit, Universal as a whole number.
You said “lateral ceph” and the transcript said “lateral staph.”
Speech models trained on general medicine mishear dentistry in specific, repeatable ways. Eva carries a correction bank built for dentistry, and most entries fire only in context: an imaging repair needs an imaging word nearby, an orthodontic one needs orthodontic words around it.
You said nothing, because it was all normal.
Dentists do not narrate normal. The finding goes unsaid, the box comes back blank, and on paper it reads as an exam you skipped. Where Eva fills a routine field she did not hear, she writes your template’s expected normal and marks it “- please verify”.
A teleconsult that “found the soft tissues within normal limits.”
When the visit reads as remote, Eva says so on the note and holds back the in-person examination baselines: findings, soft tissue, perio, occlusion, radiographs, TMJ. If she reads a visit wrong, a line in your notes saying an examination was performed outweighs it.
The recording is on the wrong patient.
Mother and daughter, same appointment slot, one chart open. If the finished note states an age, Eva checks it against the date of birth on that chart and asks out loud: is the date of birth wrong, or does this recording belong to someone else?
“Crowns on the uppers” came back as a bowel condition.
A medical condition heard once and nowhere else needs a second signal before Eva lets it stand: the same term in your typed notes, a repeat later in the consult, or medical wording around it. Without one she flags it and points the team at the intake form.
The limits
What Eva does not do
Every capability above has an edge, and a few are deliberate refusals. They are gathered here rather than hung off each card, because a page that caveats every sentence reads as hedging and gets skimmed. Nothing is softened by being collected.
She will not convert your tooth numbers until you tell her your system
No default, no guess. A brand-new clinic gets Eva declining to translate, not Eva picking the convention of whichever country wrote the most training data.
She fills some of the routine normals, not all of them
On a full examination template Eva writes the expected normal into some routine fields and leaves others blank. She also cannot tell whether a normal she wrote is true for this patient. That is what the “- please verify” marker is for: it sends you to look.
Whether a visit was remote is Eva’s read, not a stored fact
She writes it on the note and asks you to confirm it. If she reads a visit wrong, a line in your notes saying an examination was performed is enough to correct her.
She raises a flag; she never blocks the note
An age that does not match the chart, or a medical condition mentioned once and nowhere else, gets flagged for you to look at. Eva cannot tell a genuine one-off from a mishearing, so a real condition mentioned once is flagged exactly like an error would be.
She will not invent a barrier when there isn’t one
A patient who understands the plan and is not bothered enough to start yet is recorded as exactly that. There is no “not interested” barrier, deliberately. Inventing one is how a team ends up solving a problem the patient does not have.
She will not read “let me think about it” as a money problem
Not without something else in the room saying so: a voiced cost concern, a payment question, a request to stage the work. Assuming cost is how a patient who never understood the treatment gets offered a payment plan.
After the note
The part that runs after the patient says yes
This is the case layer: built and running, and being validated before wider release. It carries that tag here for the same reason every capability on this site carries the band it is actually in.
In testing
The next-day call is already on the list
Sign off a treatment note and the welfare-call task is waiting the next day, and Eva knows which visits earn one. An emergency visit gets the call even on an exam template; an impression or scan visit does not. It is a task on your team’s list, not an appointment in your book.
In testing
An escalation belongs to the dentist
When a post-op call is recorded as needing the dentist, the case moves to a dentist-owned lane, and neither snoozing the case nor logging another contact clears it. Only the dentist's answer does.
In testing
An open lab item keeps the case open
Eva does not manage your lab. She treats an outstanding lab item as a reason the case is not finished, and a temporary crown counts as not finished. A case waiting on a lab cannot quietly read as complete while the patient is still walking around in a temporary.
Eva creates the task and keeps the case visible. A person makes the call, books the appointment, and decides what happens next. None of this reaches your patients on its own.
Who Eva is for
Where Eva earns her keep
Eva is for clinics where treatment gets diagnosed and then has to be carried: restorative, implants, orthodontics, full-arch. Work that spans visits, depends on a lab, and dies quietly when one appointment in the middle goes missing.
A single-visit practice uses less of the carrying and all of the rest. The note, the team brief and the patient assessment work the same on any visit. An emergency still earns its next-day call, and the same numbers still reach your dashboard. Eva is built for the sequence; she does not need one to be worth having.
If you work across more than one practice, the sequence is the part you can least afford to hold in your head. You diagnosed it on Tuesday at one address and you are somewhere else on Wednesday; the case does not travel with you and nobody at either place is carrying it on your behalf. Eva is not tied to a building, and she is as useful to a dentist working single-handed or moving between practices as to a practice with a full team.
See it on a real consult
Book a walkthrough and bring a case of your own. The quickest way to judge any of this is to watch it run on dentistry you recognise.