Charting, perio, notes, claims, and denials in one AI-native platform. AI drafts are confirmed by clinicians, and every chart change is written to an audit trail.
Built and working — all data synthetic today. Pilot practices onboarding fall 2026.
See it work
A guided tour of the real product in the Dim theme — 18 screens end to end. Play the quick highlights or the full tour on the demo page, and turn on 🔊 for narration — real screens, synthetic data, never behind an email gate.
What changes for your practice
Dictate exams, perio probing depths, and visit notes — speech becomes structured, confirmable clinical data, not a transcript blob. A permission-aware command palette keeps navigation at your fingertips: ask in your own words, in your own language, and it takes you there.
Eligibility checked before the visit, claims scrubbed before submission, denials analyzed by AI for patterns, and appeal letters drafted for your biller's review. Privacy-preserving cross-practice learning — k-anonymity enforced in code — grows as pilot practices join.
Schedule → chart → claim → payment → recall → DSO reporting. One record, one login. It runs in the browser — nothing to install.
One platform
From the front desk to the operatory to the billing office — every step writes to the same patient record. No exports, no re-keying, no second system.
An anatomical odontogram plus six-site perio probing with a voice cursor and deterministic AAP staging that never assumes a patient is healthy.
Structured exams, an ambient scribe that drafts the visit note for you to review, and signed notes that are immutable — corrections are addenda.
An in-browser DICOM viewer where findings enter the chart only when a clinician records or confirms them — no autonomous radiograph diagnosis, by design.
Diagnosis-first planning grounded in real fee schedules, with conservative estimates that say "unknown" rather than guess a fee, and a patient-facing presentation — every number an estimate, never a guarantee.
Eligibility before the visit, claims scrubbed before submission, ERA posting, denial intelligence, and statements — the whole loop in one place.
Day-to-year calendar with family booking and recall lists, plus 19 report types that roll up across locations for groups and DSOs.
The difference that matters
AI in a medical record has to earn its place. In DzentAI, AI proposes; the clinician confirms; every chart change is audited. That isn't a policy — it's how the system is built.
AI output lands as a proposal. A clinician confirms it — or rejects it with a reason. Both outcomes are recorded.
Every suggested code is checked against the catalog on the server. Out-of-catalog answers are rejected automatically.
Once a clinical note is signed it can never be edited — corrections are addenda, on the record, in order.
Every change to a chart is appended to its history with a source tag — manual entry, voice, imported file, or AI.
Production deploys are blocked unless all backend tests and a browser-run patient-safety suite pass. There is no bypass flag in the deploy workflow.
168 database tables across 3 isolated planes: your patients' records, de-identified learning, and clinical terminology.
Cross-practice insights surface only patterns seen at 2+ practices with 3+ samples — enforced in code, not in policy.
A clinical-terminology engine with 107,862 codes ingested, with licensed-source connectors — and a guard that refuses to load licensed code sets until the license is on record.
We deliberately kept regulated diagnostic-imaging AI out of v1. Imaging findings enter the record only when a clinician records or confirms them — and where AI vision assists the doctor's chart evaluation, its output is advisory and clinician-confirmed. We build for the regulatory reality of medicine.
Medical alerts, perio staging, and exam gating are deterministic rule systems — not model output. AI drafts; rules and clinicians decide.
For your team
Built for your kind of practice
The same governed record adapts from a single startup operatory to a multi-location group.
One system instead of five subscriptions — charting, billing, scheduling, and recall in a single login. Built-in help and guided tours on 96 screens get a small team productive fast.
See how →Multi-location rollups of AR aging, collections, production vs. goals, and provider comparison across 19 reports. Corporate roles see aggregates; per-patient access is a justified, time-boxed, audited elevation.
See how →A governed clinical record — perio, imaging, structured exams, and diagnosis-first treatment planning — on a deterministic safety spine. Autonomous radiograph diagnosis is deliberately excluded from v1 by regulatory design; imaging AI contributes only advisory, clinician-confirmed input.
See how →Family consolidated booking, recall determination that never assumes a patient is healthy, and patient-facing treatment presentations — the day-to-day of a busy general practice.
See how →By the numbers
Counts from the codebase and test runs as of July 2026. Codes ingested: ICD-10-CM, HCPCS, openFDA, with licensed-source connectors and a fail-closed license guard for CDT/CPT.
An honest comparison
| Capability | Legacy desktop PMS | Cloud PMS | AI point tools | DzentAI |
|---|---|---|---|---|
| AI-native system of record | No — AI bolted on, if at all | Typically add-on modules | Not a system of record | Designed around AI from the first line |
| Clinician-confirmed AI writes | — | Varies by add-on | Varies | Propose → confirm → audit, enforced on the server |
| Deploy-blocking patient-safety test suite | — | — | — | Every production release |
| Unified clinical + billing record | Often separate modules | Often separate products | No | One record, one login |
| Privacy-preserved network learning | No pooling | No pooling | Sometimes, opaque | K-anonymity enforced in code (2+ practices, 3+ samples) |
| DSO rollups included | Extra modules | Higher tiers | No | Included — 19 report types, org totals |
Rows are limited to what we can defend from our own codebase; competitor columns describe the categories, not any specific vendor.
What practices say
We're onboarding pilot practices in fall 2026. When they've run DzentAI in their own offices, their words go here — not before. We'd rather show you the working product than invent a quote.
Want one of these to be your words? Become a pilot practice.
Pilot program
We're onboarding a small group of practices for fall 2026. Pilots work directly with the people who built it, shape the roadmap, and get pilot pricing.
Prefer email? Write to us here.
Questions we get
Pilot pricing is per practice, per month, and set with each pilot practice — talk to us. There's no per-seat math and no charge to evaluate the demo environment.
Records from previous providers import today — PDF, DOCX, and plain-text documents are structured by AI into the chart for your review before anything is saved. Full-practice migration is scoped individually with each pilot practice, and DzentAI can run alongside your current system while you transition.
All data in the product today is synthetic — no real patient data has been processed. Before any practice goes live, our checklist requires executed BAAs with our cloud and AI providers and commercial code-set licenses. The security architecture — isolated data planes, role-based access, MFA, and an append-only audit log — is already built and is described on the security page.
A clinician rejects the proposal with a reason, and the rejection is audited too. AI output only ever lands as a proposal — nothing routes into the record until a human confirms it, and suggested codes outside the catalog are rejected by the server automatically.
A modern desktop browser — that's it. Nothing to install, nothing to patch. It's designed for desktop and landscape iPad use at the front desk and chairside.
Pilot practices onboard fall 2026. Request pilot details and we'll schedule a walkthrough.