NexusMD.ai

The product reads six classes of hospital source. Its multi modal data processing covers "voice, handwriting, scanned paper, imaging, pathology, and EMR ingestion", "unified into structured outputs". All six are inbound. NexusMD calls itself "API-ready for integration into existing hospital systems and workflows. No system replacement required", under the heading "Designed to integrate".

It describes no write into any hospital system, such as a note written into the EMR or an updated coding record. No EMR product is named, whether Epic, Cerner or an Australian system, and NexusMD does not say whether connectors are native or partner built or how credentials are scoped and revoked. NexusMD has announced a partnership with RFDS Victoria for pre hospital care.

Sourcenexusmd.ai/product.html items 06 and 09 and hero; published Integrations & Tool Calling barread 2026-09-14

A fixed, linear sequence runs from capture to coding, with a human approval in between. Material from voice, handwriting, scanned paper, imaging, pathology and EMR sources is unified into structured outputs. Those become structured notes and summaries, and those become coder ready summaries aligned to the Australian classifications ICD-10-AM, ACHI, AR-DRG and MBS. Capture, structuring, clinical documentation and coding are separate stages.

The flow has no branch that NexusMD describes for an ambiguous code, no retry when capture fails and no fallback for an illegible source. It names no workflow surface that mixes deterministic steps with agent steps, and no versioning or debugging of the flow. The four care settings are environments the product suits, not configurations a customer composes. NexusMD does not say whether scribing and coding are two agents handing off or one pipeline with two outputs.

Sourcenexusmd.ai/product.html items 04, 05 and 06; published Workflow Orchestration barread 2026-09-14

Each encounter draws on the hospital's own clinical material, through voice, handwriting, scanned paper, imaging, pathology and EMR ingestion unified into structured outputs. The product reads and reasons over that material to produce a note and a code set. Material is ingested for one encounter and resolved into an output.

NexusMD names no corpus a hospital adds to, refreshes or scopes by team, no index or permissions over it, and no retrieval augmented generation. Coding is aligned to the public Australian classifications ICD-10-AM, ACHI, AR-DRG and MBS. NexusMD describes no view of which sources influenced an output, and says nothing about stale, duplicated or conflicting documents.

Sourcenexusmd.ai/product.html items 05 and 06; published Knowledge Grounding & RAG barread 2026-09-14

A clinician reviews, edits and approves every generated note, and a coder confirms every code set, before anything is finalized. There is no autonomous mode. The documentation capability covers "ambient voice capture, structured notes and summaries, with clinician review, edit and approval embedded in real clinical workflows". Approval sits between generation and finalization on every note and every code set, and nothing is committed to the record on the agent's own authority.

NexusMD describes its AI as "supporting clinicians, paramedics, coding and administrative teams, with human oversight embedded at every step". The coding output is "coder-ready structured summaries", meant to be confirmed by a coder, not to replace one. The product hands a person a draft and waits. NexusMD does not say how approval requests reach existing channels or ticketing systems, whether low and high risk actions get different autonomy, or why a checkpoint fired.

Sourcenexusmd.ai/product.html items 04 and 05, nexusmd.ai/ai.html; published Human Oversight & Guardrails barread 2026-09-14

NexusMD lists "ISO 27001 certified, HIPAA-compliant, and APP-aligned governance" under the heading "Secure & compliant by design", as one of the product's numbered feature tiles. It has no footer badge, trust portal, security page, compliance page or documentation host, and its site runs to four pages.

It names no single sign on, identity provider, provisioning standard, role based access, permission model or least privilege tool access for users, teams, tools, knowledge sources or model usage. It publishes no data retention or model training policy for the recorded clinical conversations it ingests.

Sourcenexusmd.ai/product.html items 07 and 08, site navigation read in full; published Security, Identity & Governance barread 2026-09-14

What a person reviews is the note and code set the product produces, which a clinician edits and a coder confirms. NexusMD names no audit log, run history or runtime trace. It describes no way to see how a note was derived from an encounter, which source documents fed a code, or what the system did between capture and output, and it does not address export to monitoring or SIEM systems or retention by plan. The "audit-ready coding" it promises means a code set fit for a funding or compliance audit of the hospital.

Sourcenexusmd.ai/product.html read in full, nexusmd.ai/ai.html; published Observability & Auditability barread 2026-09-14

Ambient capture holds a consultation in flight while it produces a note, as working state for that one run. The durable artifact is the note written into the hospital's record, and continuity of care comes through those structured outputs, not a store NexusMD keeps. EMR ingestion reads the patient's existing record as input for each encounter. NexusMD describes no session, conversation, workflow or long term memory, no state that survives an encounter and no context carried from one scribe run to a later one, with no scope, lifetime or purge path for anything held.

Sourcenexusmd.ai/product.html items 04 and 06, nexusmd.ai/ai.html; published Memory & State Persistence barread 2026-09-14

Processing stays 100% onshore in Australia. Under the heading Australian data sovereignty, NexusMD promises trusted cloud infrastructure with 100% onshore processing, which pins data storage and processing to one country. It publishes no region list and names no way for a customer to choose one. Trusted cloud infrastructure is the only deployment mode named, with no SaaS tier, dedicated cloud, VPC, on premises, hybrid or local runtime option. NexusMD does not say who handles upgrades and backups.

Sourcenexusmd.ai/product.html items 07, 08 and the hero; published Deployment & Data Residency barread 2026-09-14

Clinical scribing and coding come ready made for Australian healthcare. NexusMD calls the product a "hospital-grade AI layer for Australian healthcare", "tailored for Australian healthcare", with purpose built clinical agents for scribing and coding, capabilities aligned to Australian classifications and four named care settings, from pre hospital and ambulance to inpatient and outpatient. Customers get a product built for their sector, not a blank platform.

NexusMD names no browsable set of agents or templates to select and adapt, and describes no template, blueprint or configuration surface, how editable the agents are, or whether they are production ready starting points. The nine capabilities on the product page are numbered feature tiles with no per agent pages. Scribing and coding are consecutive stages of one pipeline, since the coder ready summary is produced from the structured note.

Sourcenexusmd.ai/product.html items 01 to 06 and hero, site navigation; published Prebuilt Agents, Templates & Packs barread 2026-09-14

Work begins when a clinical conversation does. A clinician starts a consultation and the system listens through "ambient voice capture", with "ambient listening and AI voice assistant tools" that "capture clinical conversations in real time, across multiple languages". The consultation itself is the channel, and the clinician does not operate an interface to start the work. That holds in pre hospital and ambulance work too, where NexusMD promises "real-time capture of clinical information under complex and time-pressure environments".

Work also arrives as documents, since handwriting, scans, pathology and other material enter alongside speech, but NexusMD does not say whether that ingestion is scheduled, watched or fed by hand. It names no schedule, webhook, message, inbox, CRM event or queue, and does not address several triggers feeding one workflow or duplicate and concurrent capture when encounters overlap in an emergency department.

Sourcenexusmd.ai/product.html items 01, 04 and 06, nexusmd.ai/ai.html; published Triggers & Channel Coverage barread 2026-09-14

The product is a layer, not a model, which NexusMD calls a "secure, compliant AI layer with data sovereignty maintained" and a "hospital-grade AI layer for Australian healthcare". It names no model or provider, and model selection is its own and undisclosed. NexusMD names no model selector, disclosed routing or bring your own key path, and publishes no benchmarking or fallback policy, trust center or subprocessor list. Inference runs onshore, since all processing is 100% onshore.

Sourcenexusmd.ai/product.html read in full, nexusmd.ai/ai.html; published Model Flexibility & Routing barread 2026-09-14

NexusMD says under "Supports system integration" that it is "API-ready for integration into existing hospital systems and workflows", with no system replacement required. That describes NexusMD fitting into a hospital's existing systems. The product is used through its own interface, plus whatever is arranged in an implementation.

NexusMD publishes no API, API reference, SDK, MCP server, endpoint, authentication model, versioning statement or documentation host, and its six item site navigation has no developers entry. It does not mention MCP in either direction, names no tool surface, and describes no fit with CI/CD, internal portals or other agent systems.

Sourcenexusmd.ai/product.html item 09, site navigation read in full; published APIs, SDKs & MCP Extensibility barread 2026-09-14

In what NexusMD calls "Australia's first real-world AI-scribe study", run in the Emergency Department of St Vincent's Hospital "using objective, analytics-driven methods", the company reports that NexusMD.ai cut ED clinicians' documentation time by an average of 56%. The study measured clinician time after deployment, as a one off measurement.

It does not say whether the notes or codes were right, how often clinicians corrected them, or whether accuracy holds across specialties, and NexusMD publishes no coding accuracy figure. It describes no testing against datasets before production, no quality gates, no optimization loop after deployment and no way to test a changed model before it reaches a live emergency department.

Sourcenexusmd.ai/ home page study section; published Testing, Debugging & Optimization barread 2026-09-14

The path runs from capture and ingestion at one end (voice, handwriting, scans and records) to a structured note and code set at the other.

NexusMD's "Built-in handwriting recognition, image-to-text conversion, and AI-assisted drug charting transform unstructured inputs into structured clinical information", so a photographed drug chart or a handwritten observation form is read as a file. NexusMD describes no browser or computer use and no screen driven at any point.

Its stated approach is to be "API-ready for integration into existing hospital systems", with "no system replacement required" for the hospital.

Sourcenexusmd.ai/product.html items 06 and 09, nexusmd.ai/ai.html; published Browser & Computer Use barread 2026-09-14