GLACESCRIBE | SPECIALTY ADAPTATION
A Cardiology Draft Should Not Read Like a Peds Draft.
GlaceScribe is tuned per specialty. The vocabulary the model expects, the way the SOAP note is shaped, the ICD-10 and CPT codes the suggestion pane surfaces, and the small details a physician in your field actually documents are configured differently for cardiology than for pediatrics, for behavioral health than for podiatry, for oncology than for internal medicine. The draft you review on a Tuesday in heart-failure clinic and the draft you review on a Wednesday in a well-child visit are not the same draft. They were never supposed to be.
WHY A GENERIC AMBIENT SCRIBE LOSES YOU AN HOUR A DAY
A Generic Model Treats Every Visit Like the Same Visit. Your Specialty Does Not.
A horizontal ambient scribe is built to draft a clinical note from any encounter in any field. The pitch is the same in cardiology and in dermatology, and the underlying model behind the draft is also the same. That single-model posture is a marketing virtue and a workflow problem. The model has no opinion about whether the visit you just ran was a heart-failure follow-up or a contraceptive counseling encounter, so the draft it generates is shaped to a generic average of clinical encounters. Your editing pass becomes the work of bending that generic note into your specialty’s actual structure, every encounter, every day.
The cost is hidden inside the time the tool was supposed to save you. You finished the visit faster because you did not type the note from scratch. You then spent four extra minutes per encounter restructuring the draft so the assessment matches the way your specialty actually documents the problem, surfacing the elements the payer policy expects to see for the procedural code, and adding the small details (a functional class, a tumor stage, a screening score, a growth percentile, a wound dimension) that the generic model did not know to capture because no encounter outside your specialty needs them. Multiply four minutes by a clinic day of twenty-eight visits and the hour you got back from the drafting work goes back to the editing work.
The second cost is in the coding pane. A generic scribe that suggests ICD-10 and CPT codes from the entire code set surfaces options that are not relevant to your field, and the cognitive overhead of scanning a long suggestion list for the three or four codes that actually apply to your practice is its own quiet tax. The model is trying to help by being comprehensive. The help is the wrong kind of help, because you do not need every code in the book; you need the codes your specialty bills against the visits your specialty actually runs. A scoped suggestion pane is a more honest tool for a specialist than a complete one.
GlaceScribe is tuned per specialty for the simple reason that this set of frictions is foreseeable. The vocabulary model, the note shape, the code suggestion scope, and the structured-data targets in the chart are configured around the encounter your specialty runs, and the draft that appears on your screen at the end of the visit looks like a note a colleague in your field might have written. The editing pass is then the small one it was meant to be from the beginning.
THE FIVE MECHANISMS BEHIND THE TUNING
What “Tuned Per Specialty” Actually Means Inside the Product
Specialty tuning is not a marketing wrapper around a single generic model. The configuration changes five concrete layers of the encounter pipeline, and each of the five is what produces the difference between a draft that fits your specialty and a draft that mostly does not. The five cards below name each mechanism and describe what it does inside the chart.
Specialty vocabulary
The terms your specialty uses (drug names, procedure terminology, anatomic landmarks, lab panels, device names, classification systems, screening instruments) are recognized natively. The cardiology configuration knows what an LVEF reading is, recognizes “NYHA class III” as a structured functional class rather than a free-text string, and treats a sentence about a stress echo as a specific test reference. The oncology configuration recognizes regimens by name, recognizes staging systems by their proper notation, and treats a port-access note as a procedural element. The configuration is the vocabulary the model expects to hear, narrowed to the field you practice in.
Specialty note structure
The SOAP shape changes by specialty. A behavioral-health draft surfaces the mental-status exam as a structured section with its own subfields rather than a paragraph of prose. A pediatric draft populates growth measurements, immunization status, and developmental milestones in the structured fields the chart already keeps. A podiatry draft separates wound dimensions, vascular and neurologic exam findings, and debridement detail into the discrete sections payer policy looks for. The model does not invent your note structure; it fills in the structure your specialty already documents against.
Specialty coding scope
The ICD-10 and CPT and E&M suggestion pane is narrowed to the code set your specialty actually bills against. A nephrology configuration surfaces dialysis and CKD-staging codes; an allergy configuration surfaces immunotherapy and allergy-testing codes; a behavioral-health configuration surfaces psychotherapy time codes and the screening-test codes the practice runs. Codes outside the specialty’s working set do not crowd the suggestion list. The pane stays scannable, the physician selects faster, and the cognitive overhead per encounter falls.
Provider-level refinement
Within a specialty configuration, the draft refines per provider as you use it. The abbreviations you favor, the section ordering you read most easily, the smart-phrase mappings already configured in your chart, the assessment style you write with: these settle in over the first few weeks of regular use. The model is not trained on your data outside your account; the refinement is local, lives inside your chart, and serves your editing pass. Two cardiologists in the same practice end up with drafts that differ in cadence and emphasis because the two physicians work differently, and the configuration adapts to each.
Physician review always
Specialty tuning sharpens what the model suggests. It does not change who decides. Every code that reaches a claim is a code the physician selected from the suggestion pane. Every order that posts to the orders module is an order the physician accepted. Every problem-list or medication-list change is a change the physician approved on signature. The line between draft and signature is the same in every specialty configuration, and there is no setting that lets the model finalize a note on its own behalf. The tuning makes the review faster. It does not replace it.
SPECIALTY WALKTHROUGH | INTERNAL MEDICINE
Internal Medicine: Polypharmacy, Chronic-Disease Trends, and the E&M Level That Holds Up
An internal medicine visit is rarely about one problem. The patient walks in with diabetes and hypertension and chronic kidney disease and depression and a recent fall, and the documentation has to capture which of those problems was addressed today, which is stable, which is trending, and what the plan is for each. The internal-medicine configuration in GlaceScribe is shaped around that visit. The assessment section is built around a problem-by-problem structure, with the chronic conditions pulled forward as a list rather than collapsed into a paragraph, and the discussion of each problem ties to the trended labs and vitals already in the chart.
Medication reconciliation surfaces as a structured medication change rather than a sentence about “discussed meds with patient.” When the encounter mentions starting a new agent, holding a current one, changing a dose, or stopping a medication because of a side effect, the draft surfaces the corresponding change in the medication-list suggestions pane, and your acceptance posts the change cleanly to the active medication list on signature. Polypharmacy reviews stop generating paragraphs you have to re-parse for billing; they generate structured changes you accept or reject one by one.
The E&M level is the quiet stake. A 99214 versus a 99215 versus a chronic-care management entry depends on how the assessment captures the number and severity of problems addressed, the data reviewed, and the risk of complications, and the documentation in front of the auditor has to match the level on the claim. The internal-medicine draft surfaces the elements the level depends on with the supporting documentation highlighted, so the physician selecting the code is looking at the same evidence the auditor will. The level you choose is the level the chart supports, every encounter, without the extra review pass at the end of clinic.
SPECIALTY WALKTHROUGH | CARDIOLOGY
Cardiology: Functional Class, Echo References, and Device-Indication Language
The cardiology visit runs on its own structured language. NYHA functional class, ejection fraction trends, stress test results, device interrogation findings, and the precise indication for the procedure or therapy being considered are the elements the chart and the claim and the consulting referral all need to see in the right place. A generic draft folds those elements into prose; the cardiology configuration in GlaceScribe surfaces them as discrete structured fields the chart already keeps for them.
A heart-failure follow-up encounter generates a draft where the LVEF reading from the most recent echo populates the assessment alongside the current functional class, the volume-status exam findings populate the physical-exam section in their proper subsections, the diuretic adjustment surfaces as a structured medication change, and the indication for an ICD or CRT evaluation (if discussed) populates with the supporting clinical language the device-implant referral will need. The encounter you ran is captured in the shape your sub-specialty actually documents against.
The coding pane is narrowed to the cardiology working set. E&M, the cardiology-specific procedural CPT codes for in-office testing and device interrogation, and the ICD-10 codes for the cardiac conditions your panel actually carries are surfaced; the long tail of codes outside cardiology does not crowd the list. Your selection happens faster, and the prior-authorization conversation that often follows a cardiology visit has the documentation it needs from the structured note rather than from a follow-up phone call to the patient.
SPECIALTY WALKTHROUGH | ONCOLOGY
Oncology: Staging Notation, Regimen Documentation, and the Infusion Suite
Oncology documentation carries more weight per word than most clinical fields. The staging notation, the molecular markers, the regimen by name, the cycle number, the dose and dose modifications, the toxicity grading, and the response assessment are all elements that have to be captured precisely and that flow into the chemotherapy authorization, the infusion-suite scheduling, the J-code billing, and the eventual response evaluation. The oncology configuration in GlaceScribe is built around that documentation density.
A treatment-planning encounter generates a draft where the cancer type and stage populate the active oncology problem, the molecular markers populate as discrete data points rather than free-text strings, the regimen surfaces as a structured plan with the cycle, the day, the agents, and the doses each in their own field, and the toxicity assessment from the prior cycle populates with the CTCAE grading the practice uses. The infusion-suite order set the encounter implies is surfaced as a suggestion the physician accepts, and the pre-medication, hydration, and growth-factor support that goes with the regimen populates with it.
The coding pane is narrowed to oncology. Evaluation and management codes appropriate to oncology visits, the chemotherapy administration CPT codes, the J-codes for the drugs administered, and the cancer-specific ICD-10 codes are surfaced; the supporting documentation for each code suggestion is highlighted in the note. Hospice and palliative-care elements, when the encounter discusses goals-of-care or symptom management, populate the appropriate fields and surface the corresponding codes without requiring a separate documentation pass.
SPECIALTY WALKTHROUGH | BEHAVIORAL HEALTH
Behavioral Health: Mental-Status Structure, Safety Screening, and the Time-Based Code
A behavioral-health encounter is the kind of clinical work where the conversation is the visit. The therapeutic content runs longer than the average encounter, the structure of the documentation is qualitatively different from a procedural note, and the elements payers expect to see (the screening scores, the safety-risk assessment, the diagnostic formulation, the therapy modality, the time spent in psychotherapy versus medication management) are scattered across what is otherwise a narrative. The behavioral-health configuration shapes the draft around that reality.
The mental-status exam surfaces as a structured section with its own subfields, not a paragraph of prose. Safety screening (suicidal ideation, homicidal ideation, self-harm, violence risk) populates as a separate documented item that the chart can carry forward across visits and that the payer can read against the medical-necessity criteria for the level of care. PHQ-9 and GAD-7 scores, when administered, post as structured data tied to the visit. The therapeutic content of the session is captured as a narrative the way you would have written it, but the elements that have to be discrete are discrete.
The time-based psychotherapy codes (90832, 90834, 90837 and the corresponding add-on codes for medication management) sit at the front of the coding pane with the documented session time highlighted. The diagnostic code suggestions are narrowed to the DSM-aligned ICD-10 set behavioral health bills against. The split between psychotherapy and evaluation and management time, when the encounter combined both, is presented with the supporting documentation for each so the physician selecting the codes is selecting against the same evidence the audit would.
SPECIALTY WALKTHROUGH | PEDIATRICS
Pediatrics: Growth Markers, Immunization State, and the Three-Way Conversation
A pediatric visit is a three-way conversation. The parent is in the room with the child, the questions and answers move between them, and the documentation has to capture both the history reported by the parent and the findings the physician took directly from the child. The pediatric configuration handles the multi-speaker audio cleanly, attributes the relevant portions of the history correctly in the HPI, and preserves the structure pediatric encounters actually run on.
A well-child visit generates a draft where growth measurements (height, weight, head circumference, BMI percentile by age) populate the structured fields the chart already keeps for them, immunization status surfaces with the doses given today and the doses due at the next visit, and developmental milestones populate as discrete items tied to the age-appropriate screening tool the practice uses. The anticipatory guidance content of the visit is captured as a narrative, and the screening for social determinants and parental concerns is captured as structured items rather than buried in prose.
An acute pediatric visit, by contrast, generates a draft built around the chief complaint, the symptom timeline, the exam findings, and the assessment-and-plan for the acute problem, without forcing the well-child-style scaffolding the encounter did not run. The coding pane carries the pediatric-specific E&M and preventive-medicine codes, the vaccine administration codes, and the developmental screening codes the practice bills against. The visit type set on the schedule drives which draft shape applies, and the configuration adjusts without a configuration step on the physician’s part.
SPECIALTY WALKTHROUGH | PODIATRY
Podiatry: Wound Dimensions, Vascular Findings, and the Procedure Code That Pays
Podiatry documentation lives close to the procedure pane. A diabetic-foot encounter, a wound-care visit, a nail-debridement appointment, and a routine biomechanical visit each carry their own structured documentation requirements, and the documentation has to match the procedural CPT code on the claim with the specificity the payer policy expects. The podiatry configuration in GlaceScribe is built around that match.
A wound-care visit generates a draft where the wound location, the dimensions (length, width, depth), the appearance of the wound bed, the periwound condition, the drainage character, and the vascular and neurologic exam findings populate the structured fields the chart keeps. The debridement detail (the tissue removed, the depth of debridement, the instrument used, the patient response) populates the procedure section in the shape the corresponding CPT code requires for documentation support. The Wagner or University of Texas wound classification, when documented, posts as structured data tied to the encounter and trended across visits.
The coding pane is scoped to podiatry: nail-debridement codes with the laterality and the number of nails, wound-debridement codes with the depth and the surface area, the diabetic-foot exam code, the routine foot-care codes with the systemic-disease modifiers the payer requires, and the orthotic and DME codes the practice provides. The supporting documentation for each suggested code is highlighted in the note, so the physician selecting the procedure code is selecting against the documentation that supports it. The denial patterns common to podiatry, the ones that come from a missing modifier or an unsupported diagnosis pairing, are surfaced as flags before the claim leaves the office.
REFINEMENT, REVIEW, AND THE FULL SPECIALTY LIST
The Configuration Refines to You, and the Physician Reviews Every Encounter
Within whichever specialty configuration is active for your encounter, the draft refines per provider as you use it. The phrasing you favor for an assessment, the section ordering you scan most quickly, the abbreviations your panel actually uses, and the smart-phrase mappings you have already set up inside the chart settle into the draft over the first few weeks of regular operation. The refinement is local to your account; the underlying AI vendors do not train on your audio or your PHI under the business associate agreements the platform holds with them. The configuration that adapts to you is yours, lives in your chart, and is not part of any vendor’s training pipeline.
The physician-review loop is the same in every specialty configuration. The draft surfaces, the order suggestions surface, the ICD-10 and CPT and E&M suggestions surface, and the physician reviews. The codes that reach the claim are the codes the physician selected. The orders that post to the orders module are the orders the physician accepted. The problem-list and medication-list changes that propagate at signature are the changes the physician approved. The specialty tuning narrows the suggestion list so the review is faster; it does not move the signature off the physician. There is no setting that lets the model finalize a note unattended, and there is no batch path that bypasses the review.
Twenty-plus specialties carry their own tuning today: internal medicine, family medicine, pediatrics, cardiology, oncology, behavioral health, podiatry, OB-GYN, orthopedics, dermatology, rheumatology, gastroenterology, nephrology, endocrinology, pain medicine, allergy, pulmonary, neurology, sleep medicine, geriatrics, urgent care, PT/OT, and skilled-nursing settings. Each configuration is on by default for any specialty you turn on at the practice level, and the model applies the configuration based on the specialty set on the encounter or the provider’s default. If your specialty is not on the list, the named account team can confirm whether the tuning is in active development or whether a generic clinical configuration is the right starting point for the encounters your practice actually runs.
The per-specialty hub at /specialties carries the practice-side documentation for how the platform fits each field, including the medical-billing posture, the practice-management workflow, and the testimonials from physicians already running on the platform in that specialty. The parent product page at /glacescribe covers the ambient documentation product as a whole: how the encounter workflow runs, what lands in the chart, how audio and consent are handled, and the bundled-not-standalone product architecture that puts the model and the chart inside one product.
FREQUENTLY ASKED
Questions Specialty Physicians Ask About the Tuning
Is the specialty configuration a separate model per specialty, or is it the same model with a prompt?
The specialty configuration is a real configuration of the encounter pipeline, not a prompt sleeve over a single generic model. The vocabulary the speech layer expects, the structured-data targets the draft populates, the suggested ICD-10 and CPT code scope, and the visit-type templates that drive the note shape are each set per specialty. The underlying AI infrastructure (transcription, language model, drafting layer) is multi-vendor and HIPAA-compliant, and the specialty tuning sits as a configuration layer on top of that infrastructure so the draft you read at the end of the encounter is shaped to your field rather than to a generic clinical average.
What happens if my practice runs more than one specialty?
The configuration applies per encounter, not per practice. A multi-specialty practice configures each specialty at the practice level, and the model applies the right configuration based on the specialty set on the encounter or the provider’s default. A primary-care practice that also runs a behavioral-health line, a cardiology consult clinic that also handles general internal medicine follow-ups, and a multi-specialty group where each provider has a different specialty assignment all run cleanly. The provider does not pick the configuration; the schedule and the provider profile pick it, and the configuration applies in the background.
My specialty is not on the supported list. What does that mean for us?
The supported list covers twenty-plus specialties today and continues to expand. If your specialty is not currently on the list, the named account team can confirm whether the tuning is in active development (with a target date), whether a generic clinical configuration is the right starting point for your encounters, or whether the practice would be a good fit to participate in the development of a new specialty configuration. The conversation is concrete: the team is honest about what is and is not ready, because the practice that turns ambient documentation on for a specialty the model is not tuned for will not get the experience the supported specialties get.
Does the model learn from my encounters in a way that benefits other practices?
No. The refinement that happens during regular use of the encounter workflow is local to your account; the underlying AI vendors are prohibited under their business associate agreements with us from training on your audio or your PHI. The configuration that adapts to your phrasing, your section preferences, and your specialty house style is yours, lives inside your account, and does not leave to benefit any other practice or any vendor’s general-purpose model.
Does specialty tuning change the review and signature step?
No. The review and signature step is the same in every specialty configuration. The draft surfaces, the order and code suggestions surface, the physician reviews, and the physician signs. The codes that reach the claim are the codes the physician selected; the orders that post are the orders the physician accepted; the problem-list and medication-list changes that propagate at signature are the changes the physician approved. Specialty tuning narrows the suggestion list so the review is faster, but the line between draft and signature is the same line. There is no setting that lets the model finalize a note on its own behalf in any specialty configuration.
How long does it take for the configuration to refine to my specific style?
The specialty configuration is in place from the first encounter; the per-provider refinement (your phrasing, your section ordering, your abbreviations, your smart-phrase mappings) settles in over the first two to three weeks of regular use. The first week is typically a set of short tuning calls with your named account team, where the providers run a few encounters with the platform and the account lead helps refine the section preferences and the smart-phrase mapping so the drafts land closer to your practice’s house style. By the second or third week, the editing pass is at its steady state.
See What Glenwood Can Do for Your Practice
A 20-minute working call. We show you the platform on your specialty’s actual workflows, not a generic demo.