Dermatology EHR With AI: What Is Real for a Small Clinic

What is real today: ambient scribes that draft the visit note (ModMed Scribe inside EMA, or an add-on such as DeepCura), code suggestions you approve, and photo or body-map tools that organise images. What is marketing: any EHR claiming to diagnose skin cancer from photos. Regulated lesion AI exists, but as separate medical devices, not EHR features.

The second filter is the phrase "coming soon". EZDERM's home page, for example, lists an AI encounter summary and AI insurance-card recognition, both marked as coming soon. A feature you cannot watch working on one of your own recorded visits during a demo is a roadmap item, and should be priced as nothing.

Follow me on Instagram@sagnikteaches

Why dermatology charting is a good test for AI

A single skin-check visit can cover a dozen anatomical sites, a cryotherapy session, a biopsy and a chronic condition such as psoriasis. Each finding needs a site, a side, a size, a description and, for procedures, depth and closure. Point-and-click templates handle this with menus and body diagrams, and the clicking is where the evening charting comes from.

Connect on LinkedInSagnik Bhattacharya

That makes dermatology both promising and unforgiving for ambient scribes. The conversation contains most of what the note needs, which is promising. But a left-right swap, a wrong lesion size or a procedure detail the clinician never said will sit in a legal record, which is unforgiving. Every feature below should be judged on one question: does it reduce the work without adding a new kind of error you have to hunt for?

Subscribe on YouTube@sagnikteaches

Real features against marketing claims

The lesion-diagnosis row deserves a closer look, because it is where the most money is wasted. Regulated products do exist. DermaSensor describes its handheld device as an adjunctive tool for physicians who are not dermatologists, to help decide whether to refer, and says it should not be the sole diagnostic criterion. Skin Analytics describes DERM as a separately regulated device that assesses dermoscopic images and discharges low-risk patients at triage in the services that use it. Neither is a feature of a dermatology EHR, and neither is aimed at replacing a dermatologist's assessment. If an EHR salesperson implies their system diagnoses lesions, ask for the device's regulatory documents and intended-use statement; the conversation usually ends there.

The useful AI is less glamorous. An ambient scribe on a tablet records the consultation, drops the small talk and drafts the note. Whether that is worth paying for is mostly a question of accuracy on your visits and time saved per day, which is why it pays to read whether a scribe pays for itself in a private clinic and what AI scribes cost small clinics before any demo.

ModMed, Nextech, EZDERM and the add-on scribes compared

A small clinic has two routes: a dermatology-specific EHR with AI built in, or the current system plus an add-on scribe. What each vendor's own page showed on 4 November 2026:

ModMed's case is integration. Its ModMed Scribe page describes a scribe built directly into EMA that suggests note content and billing codes, with data flowing into the record once you approve. That removes the copy-and-paste step an add-on needs. The trade-off is that you are buying an EHR migration, with the contract length and data move that implies, to get it.

The add-on route keeps the current system. DeepCura publishes its prices on its pricing page, lists integrations with several named EHRs, and says team members who only read or edit do not need a paid seat. Its page does not describe dermatology-specific templates, so test its custom prompts on a full skin check. Suki is a mature ambient product, but its home page lists deep integrations with four large-hospital EHRs and no published price, so a small dermatology clinic on a specialty system should expect to copy notes across. For the patient side of either route, see introducing a scribe to patients so they stay comfortable with it.

A full skin check through an ambient scribe

An illustrative consultation, written for this tutorial, and the draft a scribe might produce from it. The point is what the clinician has to fix.

DOCTOR: "Good morning, [patient name]. Let's do your annual skin check. Any moles changing, bleeding or itching?"
PATIENT: "No bleeding, but there's a brown spot on my upper back that looks darker lately. It catches on my bra strap."
DOCTOR: "Let's look at that first. Right upper back, about four centimetres from the spine of the shoulder blade, a five-millimetre pigmented flat lesion. Under the dermatoscope there's an irregular network with some blue-grey dots and asymmetry at the top edge. I'd like to biopsy it today to rule out an atypical mole or early melanoma."
PATIENT: "Whatever you think is safest."
DOCTOR: "A three-millimetre punch biopsy, two absorbable stitches, and it goes to the lab. Across your shoulders, several seborrhoeic keratoses, the waxy stuck-on ones, which are harmless. On the left ear rim, a three-millimetre rough scaly spot: that's an actinic keratosis. I'll freeze it today, about ten seconds with liquid nitrogen."
PATIENT: "Fine, let's do both."

An illustrative draft note:

CHIEF COMPLAINT: Annual skin examination; changing pigmented lesion, right upper back.

HISTORY: Darkening pigmented lesion on right upper back, catches on clothing. Denies bleeding.

EXAMINATION:
- Right upper back: 5 mm pigmented macule, ~4 cm from scapular spine. Dermoscopy: atypical network, blue-grey dots, asymmetry superiorly.
- Shoulders, both sides: multiple seborrhoeic keratoses.
- Left ear helix: 3 mm scaly papule consistent with actinic keratosis.

PLAN:
1. Right upper back lesion: 3 mm punch biopsy under local anaesthesia (1% lidocaine with adrenaline). Closed with two 4-0 sutures. Specimen to dermatopathology. Suture review in 10-14 days.
2. Left ear helix actinic keratosis: liquid nitrogen cryotherapy, one 10-second freeze-thaw cycle. Aftercare explained.
3. Seborrhoeic keratoses: benign, reassured, no treatment.
SUGGESTED CODES: [shown by the software for the clinician or coder to confirm]

The draft reads well, which is the danger. Three details in it were never said: the anaesthetic and its strength, the suture size, and a suture review for stitches the doctor called absorbable. "Aftercare explained" was not said either. Each is the kind of plausible filler a model adds because notes usually contain it. The clinician also has to confirm the sides ("right upper back", "left ear") and the sizes (3 mm punch, 5 mm lesion), because a swap between two numbers said a sentence apart is a common scribe error. For a full checklist of what to look for, see the AI scribe note errors doctors must catch.

Migrate or add a scribe? Costing a two-dermatologist clinic

Illustrative figures throughout; replace them with your quotes. Picture two dermatologists, each working four clinic days a week for 48 weeks, each spending about an hour and three-quarters most evenings finishing notes. The clinic's current general EHR costs $200 per provider per month.

Option A, migrate to a dermatology EHR with a built-in scribe. ModMed does not publish prices, so assume a quote of $550 per provider per month including the scribe, plus $4,500 for data migration and onboarding. The clinic stops paying for the old EHR.

Option B, keep the current EHR and add DeepCura. Two seats on annual billing at $999 each, plus about $600 for two tablets and stands.

The break-even row is the useful one. Across two doctors and about 384 clinic days, Option B pays for itself with fewer than one extra visit a week. Option A needs about one extra visit every four clinic days per doctor in year one, before any value is put on the evenings the doctors get back. Neither number is a forecast; they are the bar each option has to clear.

What would make Option A win anyway: if the old EHR's charting is so slow that the scribe's output cannot be used without retyping, or if integrated code suggestions catch billing the clinic currently misses. What would sink Option B: if copying each note into the old EHR takes two minutes a visit, that is about an hour a day across the clinic, which can wipe out much of the saving. Time it during the trial. And check what DeepCura's 1,000 monthly credits per seat cover in visits for your volume; the pricing page sells an extra-credits add-on at $59 a month, which would change Option B's sum. To understand what happens to the audio during a visit, read how an AI scribe works in a medical consultation.

Biopsy tracking: the feature worth more than most AI

A lost or unreviewed biopsy result is one of the most serious failures in a dermatology clinic, and the protection is mostly ordinary software, not AI. In any demo, ask the vendor to show the whole loop on a test patient:

1. Requisition: does the order carry the site, side, size, clinical description and differential from the note, or is it typed again?
2. Labelling: does the system print a specimen label with patient identifiers and site at the chair?
3. Pending log: is there a list of every biopsy without a result, with an alert after a set number of days (for example seven working days) so someone calls the lab?
4. Result to patient: is there a task that stays open until the result has been reviewed by the clinician and the patient told?

If the scribe can feed the requisition from the dictated note, that is a genuine AI benefit, and ModMed describes approved scribe suggestions flowing into downstream workflows such as lab orders. But a clinic without a reliable pending-biopsy log has a bigger problem than charting time, and should fix that first.

Consent and recording rules for skin examinations

Recording in a room where patients undress needs more care than recording a routine follow-up. Three practical steps:

• Tell patients before the visit. A paragraph in the registration pack: the clinic uses a documentation assistant that records the consultation to draft notes, so the doctor can look at the patient rather than a screen.
• Ask again in the room. "To help me focus on you, I use a documentation assistant on this tablet that drafts my notes. Is it all right if I switch it on?" If the patient hesitates, chart manually without comment. Switch it off before any part of the examination where the patient is uncovered, if the patient prefers.
• Get the retention terms in writing. Ask how long audio and transcripts are kept, whether either is used to train models, and how deletion works. Keep the answer with your contract and check it again at renewal, because vendors change defaults. The list in security questions for AI scribe vendors is a good starting set.

Five dermatology problems scribes create and how to prevent them

1. Left and right swapped

A scribe hears words, not positions. "Right forearm, a scaly plaque, and similarly on the left an inflamed nodule" can come back with the findings on the wrong arms. The fix is a speaking habit: say each lesion as a numbered block.

"Lesion one: left forearm, back of the arm, three centimetres below the elbow. Two-millimetre red papule. Three-millimetre punch biopsy.
Lesion two: right shin, five centimetres above the ankle bone on the outside. One-centimetre plaque. Cryotherapy."

Numbered blocks make swaps much rarer and much easier to spot in review, because each block either matches what you did or it does not.

2. Code suggestions that need documentation behind them

Billing a procedure and a separate visit on the same day usually needs the note to show a separately identifiable problem. An illustrative suggestion panel:

SUGGESTED FOR REVIEW
- Procedure: punch biopsy, right upper back
- Procedure: destruction of 1 premalignant lesion, left ear
- Visit: established patient
- Note: separate visit supported? Documentation mentions full skin examination and seborrhoeic keratoses assessment. Confirm medical necessity wording.

What to do with it: the suggestion is a prompt for the coder, not a decision. If the note does not clearly describe the separate problem, the fix is in the documentation, not in accepting the code. Have your coder review the first 50 suggested claims against your payers' rules before you trust the panel.

3. Body maps and dermoscopy follow-up

A solo clinic following high-risk mole patients every six months pins each dermoscopy photo to a point on the body map. At the next visit, the previous image sits beside the new one at the same scale. EZDERM describes its 3D body map without calling it AI, and that is accurate for most products: the value is organisation and comparison, and the clinician makes the judgement. When a vendor labels this AI, ask what the AI part actually does.

4. Noise in the procedure room

Smoke evacuators, cryotherapy canisters and an assistant counting sutures all reach the microphone. Scribes sometimes turn that into nonsense findings, such as a chest or heart finding nobody examined. Two fixes: a directional or lapel microphone, and, where the product allows custom prompts (DeepCura lists them), an instruction along these lines:

This recording comes from a dermatology procedure room.
Ignore equipment noise and instrument counting.
Do not record any examination finding outside the skin unless the clinician states it explicitly.
Keep to lesion sites, sizes, descriptions, procedures, closure and aftercare that the clinician actually says.

Then check the next ten procedure notes for anything the clinician did not say.

5. Counting lesions treated with cryotherapy

Some code sets bill lesion destruction by count, with different codes for the first lesion, each additional one and larger numbers. A clinician who says "froze several on the face and arms" leaves the coder guessing and the claim usually goes in at the lowest count. Say the number:

"Cryotherapy to forehead, left temple, right cheek and the backs of both hands: eight actinic keratoses in total."

The scribe records "eight", the code suggestion counts eight, and the coder has a sentence in the note to point to if the claim is queried.

A 30-day trial plan before any EHR contract

1. Measure the baseline. For two weeks, each clinician logs charting time outside clinic hours. Without this number, no trial result means anything.
2. Trial the add-on first. Run an add-on scribe for 30 days on real visits, with consent. It is the cheapest way to learn whether ambient notes work for your clinicians' speaking style. Check it recognises terms such as "lentigo maligna", "molluscum contagiosum" and "clobetasol propionate".
3. Score every note for one week. Count edits per note and sort them: wrong side, wrong size, invented detail, missing finding. Invented details are the category that should stop a rollout.
4. Time the copy-across. If the scribe does not write into your EHR, time the paste for twenty visits. That is the hidden cost of the add-on route.
5. Only then take EHR demos. Ask each EHR vendor to run its scribe on one of your recorded visits (with consent) and compare the edit count with the add-on's.

The clinic that does this ends up paying for the features it has watched work on its own patients, and for nothing else.

Dermatology EHR and AI feature questions

Can a dermatology EHR diagnose skin cancer from photos?

No EHR reviewed here claims to. Regulated skin-lesion AI does exist, but as separate medical devices with their own intended use, such as DermaSensor, which its maker describes as an aid to referral for physicians who are not dermatologists, and Skin Analytics' DERM, a triage device. Treat any EHR sales claim of diagnosis from photos as a reason to ask for its regulatory documents.

What is the difference between a built-in scribe and an add-on scribe?

A built-in scribe such as ModMed Scribe in EMA drafts the note inside the record and can populate fields once you approve its suggestions. An add-on such as DeepCura runs beside your current system; it integrates with some named EHRs and otherwise means copying the finished note across, which costs a minute or two per visit.

How do AI code suggestions work in practice?

The software reads the finished note (sites, sizes, procedures, the separate problems discussed) and proposes diagnosis, procedure and visit codes. ModMed describes suggesting visit notes and billing codes for you to approve; DeepCura lists an automated code generator. A clinician or coder still confirms every code, because the suggestion is only as good as the documentation.

What does AI add to a dermatology clinic's software bill?

Add-on scribes publish prices: DeepCura lists $129 a month per seat on monthly billing or $999 a year. The dermatology EHRs themselves (ModMed EMA, Nextech, EZDERM) and Suki quote on request, so get the scribe priced separately in any EHR quote and ask whether it is included or extra.

Further reads

• AI Dictation Software for Doctors: Dictation or Ambient Notes? — Dictation against ambient listening, for clinicians weighing both.
• AI Medical Transcription Services vs Human Transcribers for Small Clinics — Automated transcription against human medical typists.
• AI Platform for Doctors: One Suite or Separate Scribe and Phone Tools? — One clinical suite or separate tools: the trade-offs in detail.
• Physiotherapy Clinic Software Compared: Which Has Built-In AI? — The same built-in or add-on question for another specialty clinic.
• AI Scribe for Medical Practices: Pick by Country and EHR — A decision rule for choosing a scribe by specialty.
• Best AI Scribe for Urgent Care: Fast-Visit Notes and Coding Checks — How scribes cope with short, high-volume visits.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ModMed dermatology and ModMed Scribe pages; Nextech dermatology page; EZDERM home page; DeepCura pricing page; Suki home page; DermaSensor and Skin Analytics home pages. All read 4 November 2026.