1. What is an AI-first EMR?
An AI-first EMR puts AI inside the jobs that eat a clinic’s time, and keeps a person in charge of every result. It is not a chat window bolted onto the side of a chart. The test is whether the AI does work you would otherwise do by hand: writing the visit note, typing a lab report into the chart, rebuilding forms when you switch systems, or finding the right screen when a new hire is stuck.
| The job | Without AI | With AI in Aminova |
|---|---|---|
| Visit documentation | The provider writes the note after clinic, from memory | The scribe drafts a SOAP or DAP note from the visit; the provider edits and signs |
| Lab reports that arrive as PDF or fax | Someone types each value into the chart | AI reads the report into markers, with the lab’s own ranges, for review |
| Reading a hormone or metabolic panel | The provider scans every marker for flags | A lab summary lists what is out of range, why it matters and suggested follow-ups |
| Moving from another EMR | Weeks of mapping spreadsheets and rebuilding forms | AI maps the export, rebuilds intake PDFs and places consent fields; staff approve |
| Learning the software | Support tickets and screen-share calls | Nova, the built-in assistant, answers in the app, typed or spoken |
The result for a small clinic is less admin per patient, which is the difference between hiring another coordinator and not. For the wider list of what an all-in-one system should cover beyond AI, see the 20 features most systems skip.
2. Where does AI work inside Aminova today?
Nine places today, and one on the roadmap. All of them run on the platform, so there is no separate AI subscription to buy and no API key to paste in.
| Feature | What it does | Who has the final say | Status |
|---|---|---|---|
| Nova | Answers staff questions about the software, typed or spoken, and can take you to the right screen | The person asking | Live |
| AI scribe | Drafts a SOAP or DAP note from a telehealth or in-room visit | The provider, who edits and signs | Live |
| Lab report reading | Turns a lab PDF or faxed report into structured markers with the printed ranges | Staff who review the result | Live |
| Lab summary | Summarizes a panel, explains each flagged marker and lists suggested follow-ups | The provider | Live |
| Export mapping | Works out what each file and column in your old EMR’s export holds | The clinic, who approves each suggestion | Live |
| Document filing | Reads the first page of imported documents and files labs, consents and cards into the right chart folder | Only moves what it is sure of | Live |
| Intake form rebuild | Turns a PDF intake questionnaire into a live online form | Staff edit the draft before sending | Live |
| Consent field placement | Places signature, initials, date, text and checkbox fields on a consent PDF | Staff review in the PDF designer | Live |
| Patient chat assistant | Answers patients’ logistics questions in the portal; clinical and dosing questions go to staff | Hard escalation rules run first | Live |
| Plan suggestions | Suggests a next step for every patient to raise practice efficiency | The provider | Roadmap — not shipped |
3. What is Nova, the built-in AI assistant?
Nova is the assistant your staff ask when they do not know where something is. Ask “how do I send a consent before a visit?” or “where do I start a refill?” and she answers from Aminova University, our library of step-by-step guides to every part of the product, with a link to open the guide or a button to take you to the screen.
You can type or talk to her. While a clinic is setting up and training on practice patients she floats in the corner of every screen; once the clinic is live she lives on her own Assistant page by default, so she is not hovering over real charts. She knows which page you are on and what your role is, and nothing more: she is built not to read chart text, names or field values.
Nova is the part of an AI EMR that shortens onboarding for staff rather than providers. A front-desk hire who can ask the software how it works needs fewer hours of someone else’s time.
4. What do the AI scribe, lab AI and AI import do?
The AI scribe listens to a telehealth or in-room visit and drafts the note in SOAP or DAP format. It is told that inventing a clinical fact is the worst possible failure: every drug, dose and number in the draft has to trace to something said in the visit. It biases its spelling toward the products in your own inventory, so “tirzepatide” and “sermorelin” come out right. Nothing files until the provider signs. The full detail, with TRT and GLP-1 examples, is in the scribe and lab post and on the AI scribe page.
Lab AI works in two layers. AI reads the lab report itself, copying each value, unit and printed range exactly; whether a marker is high or low is then worked out by our code from those numbers, not decided by the model. The lab summary on top is written from a fixed clinical rule set covering 38 hormone, thyroid, metabolic, lipid, blood-count and vitamin markers, so the same panel always produces the same summary.
AI import is how a clinic moves in. AI reads the column names in your old system’s export and suggests where each belongs, rebuilds PDF intake forms as live forms, and places signature boxes on consent PDFs. It is the reason our onboarding went from about three weeks to about three days. The whole process is in from 3 weeks to 3 days.
5. How does Aminova stop AI from inventing clinical facts?
By deciding in advance what the AI is allowed to produce, and making a person approve everything that reaches a chart. These are design rules in the product, not settings a clinic has to remember to switch on.
| Rule | Where it applies | What it prevents |
|---|---|---|
| AI output is a draft until a person approves it | Scribe notes, form rebuilds, consent fields, import mappings | An unread AI result becoming part of the record |
| Every number must trace to its source | Scribe drafts and lab reading | A plausible dose or value nobody said or printed |
| A missing detail stays missing | Scribe drafts and lab reading | Half a fact being completed from the model’s general knowledge |
| High and low are computed, not asked for | Lab reading | A model misjudging a result against a range |
| Column names only, never values | Import mapping | Patient data leaving the database to answer a question about structure |
| Suggest, never apply | Import mapping and document filing | A silent wrong mapping found months later |
| Style can be learned, content cannot | Scribe preferences | The scribe learning to add findings a provider usually adds by hand |
6. Is AI in an EMR HIPAA compliant?
It can be, if the AI that touches patient data runs under a signed Business Associate Agreement and the vendor limits what it sends. In Aminova, the AI that handles patient information runs on a HIPAA-covered AI model under a signed BAA, and under those terms what we send it is not used to train the model.
We also send less than we could. The scribe sends the visit transcript with age, sex and visit type, not the patient’s name, record number or date of birth. Import mapping sends column names, not rows. Nova reads which page you are on, not the chart. Where a visit is recorded, the patient’s recording consent is required first.
| Ask any AI EMR vendor | A good answer sounds like |
|---|---|
| Is the AI model covered by a signed BAA? | Yes, and they can say which agreement covers it |
| Is our data used to train the model? | No, under the provider’s terms |
| What exactly is sent to the model? | A specific, short list, not “the chart” |
| Can the AI write to the chart on its own? | No |
| Is AI use recorded in the audit log? | Yes, with the user and the patient record |
For the broader BAA question across your whole software stack, see which vendors sign a BAA.
7. What is on the AI roadmap?
Plan suggestions: AI that looks across each patient and suggests the next step, to raise practice efficiency. This is not built yet. The idea is that the EMR already holds what a provider looks at before deciding what to do next — the latest labs, where the patient is in a protocol, when a refill falls due, when they were last seen — so it can surface a suggested plan for each patient instead of waiting for someone to notice.
When it ships it will follow the same rules as everything above: a suggestion for the provider to accept, change or ignore, never an order placed on its own. We will not give a date until we can keep it.
8. How do you test an AI EMR in a demo?
Make every AI claim run on something real during the call.
| # | Ask them to | What you learn |
|---|---|---|
| 1 | Run the scribe on a short role-played TRT or GLP-1 follow-up with a dose change | Whether the dose comes back exactly as said |
| 2 | Show where the draft note goes before it is signed | Whether the AI can write to the chart on its own |
| 3 | Upload a lab PDF from your own lab | Whether values and ranges come out as printed |
| 4 | Rebuild one of your intake PDFs into a form | How much editing the draft needs |
| 5 | Place fields on one of your consent PDFs | Whether signature and date boxes land on the lines |
| 6 | Ask what is live and what is roadmap | Whether the vendor separates the two without being pushed |
If you are comparing systems more broadly, our buyer’s guide for TRT, GLP-1, peptide and longevity clinics covers the rest of the evaluation.
9. Frequently asked questions
What is the best AI EMR for a wellness clinic?
The one whose AI saves your team the most hours with the fewest risks, which you can only judge by running it on your own visits, labs and forms in a demo. Aminova is built as an AI-first EMR for cash-pay TRT, GLP-1, peptide, wellness and aesthetic clinics, and we will run every feature above on your material.
Does Aminova charge extra for AI?
No separate AI subscription. The scribe, Nova, lab reading and AI import run on the platform, with no API key to connect.
Can the AI prescribe or change a dose?
No. AI in Aminova drafts notes and reads documents. Prescribing, dose changes and signing stay with the provider.
Is our patient data used to train AI models?
No. The AI that handles patient information runs on a HIPAA-covered model under a signed BAA, and under those terms the data we send is not used for training.
Does Aminova recommend treatment plans with AI?
Not yet. Plan suggestions for every patient are on the roadmap and not shipped. Today the lab summary lists suggested follow-ups for flagged markers, and the provider decides.