1. What does an AI medical scribe do in a TRT or GLP-1 visit?
It listens to the visit and drafts the note, so the provider edits instead of writing. In Aminova the scribe works on telehealth visits and on in-room visits, where the consultation is recorded in the browser. The provider picks the note format before the visit starts, SOAP or DAP, and the draft lands in the note panel when the visit ends.
| Step | What happens | Who does it |
|---|---|---|
| Before the visit | Choose whether to use the scribe and which note format | Provider |
| Consent | The patient’s recording consent must be on file, or no note is drafted | Patient, through the clinic’s consent |
| During the visit | The conversation is transcribed | AI |
| After the visit | A draft SOAP or DAP note appears in the note panel | AI |
| Sign-off | The provider edits and signs; until then nothing is in the chart | Provider |
The scribe’s product page is AI scribe, and the step-by-step guide is in Aminova University.
2. Why do generic AI scribes get hormone and peptide visits wrong?
Two reasons: they mishear the drug names, and they tidy up the numbers. A scribe that never heard “tirzepatide” or “BPC-157” spelled correctly guesses. And a scribe tuned to write fluent notes will round, complete or convert numbers so the note reads well.
Aminova’s scribe handles the first by biasing transcription toward the vocabulary of optimization medicine and toward the products in your own inventory, which the EMR already knows. It handles the second with a rule it is given before anything else: inventing a clinical fact is the worst possible failure, and every drug, dose, unit, frequency and number in the note must trace to a line of the visit.
| Said in the visit | The draft writes | The draft never writes |
|---|---|---|
| “Still on 120 a week, Mondays” | Testosterone 120 mg weekly, injects Mondays (as stated) | A different dose, a route nobody named, or a calculated monthly total |
| “We’ll go up next month” | Discussed increasing the dose next month | The next step of a titration ladder from memory |
| “I think my last weight was around 230” | Pt reports last weight approximately 230, self-reported | A firm 230 lb, or a BMI nobody calculated |
| “Some nausea the first week, fine since” | Nausea in the first week, none reported since | “Tolerating well” with the history dropped |
| Nothing said about labs | Nothing about labs | A default lab panel or follow-up interval |
A missing detail costs the provider ten seconds to add. An invented one gets signed into the medical record. The scribe is built around that difference.
3. Who signs the AI-drafted note?
Always the provider. The scribe cannot write to the chart. The draft is handed to the provider, who edits and signs it through the same path as any other note, so the provider is the author of record.
Over time the scribe can suggest style preferences from how a provider edits its drafts: bullets instead of prose, a particular heading, shorter sections. It suggests a pattern only after seeing it at least three times, and it is forbidden from learning content. If a provider routinely adds an exam finding the visit did not contain, the scribe does not learn to add it for them.
The patient’s name, record number and date of birth are not sent with the transcript; the AI gets age, sex and visit type for context. It runs on a HIPAA-covered AI model under a signed BAA, and under those terms what we send is not used to train the model. Recordings are kept only where the patient’s consent says so, for as long as it says.
4. How does AI read a lab report?
It copies the report into the chart; it does not interpret it. When a lab result arrives as a PDF or a fax, AI reads each marker’s name, value, unit and the reference range exactly as printed. If it cannot find a value, the field comes back empty for staff to fill in rather than as a best guess.
| On the report | In the chart |
|---|---|
| Testosterone, total 286 ng/dL, range 264–916 | The same value, unit and range, as printed |
| PSA <0.1 | Recorded as below 0.1, not as 0.1 |
| A result of POSITIVE or NOT DETECTED | Kept as written, not dropped for having no number |
| No range printed | No range invented; the marker is shown as not assessed |
Whether a marker is high, low or critical is then worked out by our code from the printed numbers, not decided by the model. Clinics can also keep their own target ranges, for example a testosterone goal narrower than the lab’s range, which are shown beside the lab’s range and labeled as the clinic’s target.
If your results come through a connected lab feed rather than PDFs, see which labs work with cash-pay clinics.
5. How does AI lab interpretation work for hormone and metabolic panels?
The lab summary reads the whole panel and writes what needs attention, marker by marker, with suggested follow-ups. In Aminova it is written from a fixed clinical rule set covering 38 markers across hormones, thyroid, glucose and insulin, lipids, inflammation, blood counts, vitamins and minerals. A rule set, rather than free text from a model, means the same panel always gets the same summary, and every sentence traces to a threshold and a value.
| Marker (invented example) | Flag | What the summary explains | Suggested follow-up for the provider |
|---|---|---|---|
| Testosterone, total 250 ng/dL | Low | Below 300 ng/dL meets the usual threshold for hypogonadism; confirm whether it is primary or secondary | Order LH and FSH if not present; TRT consult |
| Estradiol 48 pg/mL on TRT | High | Suggests excess aromatization; treat only if symptomatic, and avoid over-suppression | Consider a low-dose aromatase inhibitor; recheck in 4–6 weeks |
| Hematocrit 54% | High | A primary TRT safety marker; higher viscosity and clotting risk | Dose review, hydration, phlebotomy; recheck the blood count |
| HbA1c 5.9% | High | Above 5.7% indicates prediabetes, reflecting about three months of glucose | Glycemic plan; recheck in 90 days |
| Fasting insulin, elevated | High | Insulin resistance, which often comes years before diabetes | Lifestyle protocol; recheck in 90 days |
Two rules keep it honest. A marker that is flagged but has no rule behind it is named, not skipped, so the summary never calls a panel clean because it had nothing to say about potassium. And a marker with no printed range is listed as not assessed rather than normal. The provider decides what, if anything, is released to the patient.
6. What does a TRT or GLP-1 follow-up look like with both?
Labs read before the visit, the visit drafted after it, and the provider signing one note.
| When | What happens | Where AI helps |
|---|---|---|
| Results arrive | The lab PDF or fax lands in the labs queue | AI reads it into markers with the printed ranges |
| Before the visit | The provider reviews the panel and the summary | Flagged markers and suggested follow-ups in one view |
| The visit | Dose, side effects and next steps are discussed | The scribe transcribes |
| After the visit | The note is drafted | A SOAP or DAP draft with the numbers as spoken |
| Sign-off | The provider edits, signs and acts | None; this is the provider’s step |
For what else a TRT clinic’s EMR has to handle, from controlled-substance prescribing to injection logs, see choosing an EHR for a TRT clinic and GLP-1 clinic management software.
7. Frequently asked questions
What is the best AI medical scribe for a wellness or TRT clinic?
One that gets doses and drug names right and never adds what was not said. Test it on a role-played follow-up with a dose change. Aminova’s scribe is built into the EMR, uses your own inventory as vocabulary, and produces a draft the provider signs.
Does the AI scribe work for in-person visits?
Yes. In Aminova it works on telehealth visits and on in-room visits, where the consultation is recorded in the browser. The patient’s recording consent is required either way.
Can an AI scribe add medications or doses to a note?
Aminova’s scribe is instructed that every drug, dose and number must trace to something said in the visit, and a drug named without a dose stays without a dose. The provider reviews and signs every note.
Can AI interpret testosterone and estradiol labs?
Aminova’s lab summary flags each out-of-range marker, explains why it matters for a hormone patient and lists suggested follow-ups, for example a hematocrit recheck on TRT. The provider decides what to do.
Is AI lab interpretation a diagnosis?
No. It is a summary of what is out of range and what a provider might consider next. Diagnosis and treatment stay with the provider.