Suki

Two model driven layers make up the product. Ambient capture listens to the whole visit and drafts the note, patient instructions and orders. A voice command layer interprets spoken instructions to order, navigate the chart, stage orders and query the record. That command layer is what separates Suki from a passive scribe, and it is an AI capability from end to end.

Around those two layers the homepage describes documentation, coding, clinical reasoning and question answering combined in a single solution, with voice enabled editing and problem based charting inside the note.

Voice commands and ambient capture stage orders rather than submit them, which keeps a clinician decision between the model and any clinical action. With ambient order staging, the clinician speaks the recommended orders and Suki structures and codes them. Each generated prescription order can then be accepted, edited or rejected before it goes to the EHR for review and sign off. Notes follow the same pattern, drafted for the clinician to review and confirm before signing.

A misread medication order carries a higher consequence than a flawed sentence in a note. How the system responds when a command is misheard, and how often commands are interpreted correctly, is not published.

Data is the center of Suki's case for its models. The homepage says AI is only as good as the data it is trained on, and that EHR integrations and industry partnerships give the company an extensive dataset to draw on. The developer documentation sets the terms for that data. Anything used for model training and product improvement is de identified, audio is split into isolated chunks that cannot be reassembled into the original recording, and transcripts have personal information removed.

The documentation also shows how the pipeline behaves. On the developer platform, ambient sessions detect and transcribe more than 80 languages and always produce the note in English.

Models, versions and accuracy figures are not published.

Google Cloud stores the data at rest, in services such as Cloud SQL and Google Cloud Storage, and a Cloudflare web application firewall sits in front of all edge traffic. Training draws on de identified and anonymized data, with audio broken into isolated chunks and transcripts stripped of personal information, and the homepage adds that EHR integrations and industry partnerships give Suki an extensive dataset to draw on.

The trust center carries a subprocessor page, but the foundation model provider and the other parties that process encounter content are not named publicly.

Suki's homepage claims an 81 percent increase in practice satisfaction, a 41 percent reduction in time spent taking notes, a 60 percent reduction in feelings of burnout, a 48 percent reduction in amended encounters, and an average increase of $1,688 in incremental monthly revenue per user.

Deployment scale carries most of the public case. MedStar Health expanded its partnership and put the product in front of thousands of clinicians across ambulatory specialties, alongside partnerships with Zoom and athenahealth. On athenahealth, Suki reported in January 2026 more than 450 practices and 3,400 monthly active users across Suki for Clinicians and athenahealth's Ambient Notes, about 500,000 clinical notes a month, and an adoption rate above 82 percent for Suki for Clinicians. A dedicated clinical operations team reviews note quality, a process control rather than a measured outcome.

None of the homepage figures comes with a study design, sample or period.

Encounter audio is handled narrowly. Suki does not identify the clinician or the patient in voice recordings, creates and keeps no voice signatures, and uses voice only to transcribe and summarize the note. Audio and transcripts are permanently deleted after 30 days, and the final clinical note is kept for the length of the service contract.

Training runs on de identified data. Audio used for model training and product improvement is broken into isolated chunks so the original recording cannot be rebuilt, and transcripts have personal information and protected health information removed. The homepage separately points to Suki's EHR integrations and industry partnerships as the source of an extensive dataset.

Patient consent is the customer's job. Before an integration sends personal data, the clinician is responsible for obtaining it. That is a defensible split for a platform sold through integrators, and it leaves consent to recording with the customer.

Customers get a signed business associate agreement for patient data handling, and the developer documentation states that Suki complies with HIPAA requirements. The privacy policy states directly that Suki acts as a business associate for healthcare.

Consent sits with the customer. Before an integration sends personal data to the platform, the clinician is responsible for obtaining patient consent, keeping their own consent policies and securing every necessary authorization, so consent to ambient recording is handled on the customer's side rather than by the product.

Suki's web application carries a business associate agreement page, and the developer documentation mentions the agreements without reproducing their terms.

Data moving to and from the platform is encrypted with TLS 1.2, data at rest with AES-256 in Google Cloud services, and a Cloudflare web application firewall protects all edge traffic. Run time analysis watches for anomalies and suspicious software behavior, which makes monitoring an active control rather than a static claim.

The homepage states SOC 2 Type 2 certification and HIPAA compliance, and naming the type settles the first question a SOC 2 claim raises. A dedicated trust center is published.

The SOC 2 reporting window and scope are not stated, and certification currency is not shown. The trust center may list further documents, but no ISO 27001 or HITRUST certification, penetration testing statement or subprocessor list is published.

Suki positions its outputs as decision support, and clinicians review and confirm notes before signing, so the signature is the control point. Suki holds and claims no FDA clearance, and none is required for ambient documentation.

Section 5.4 of the terms disclaims any guarantee that output from the AI features is accurate or complete, places final review on the customer, and states that the platform is not an electronic medical record system. The only affirmative statement of scope is a definition describing the platform as providing medical transcription and related services.

The product has moved well past transcription. It offers ICD-10, E/M and HCC coding assistance, answers questions against the specific patient record, takes voice commands for ordering and charting, and stages orders, starting with prescription orders in athenahealth; a staged order puts a proposed clinical step into the chart. How those functions relate to FDA rules is not published.

Audio quality is the limit Suki acknowledges. The company states that transcription and predictive performance depend on the quality and completeness of the audio, and that noisy environments or poor microphones reduce accuracy. That is an operational caveat rather than a fairness disclosure, but it points at the same exposure. Speech recognition accuracy varies with accent, dialect, speech rate and vocal characteristics, and ambient sessions on the developer platform accept more than 80 languages, each producing an English note.

The public material is substantial on security and compliance and silent on AI governance. No performance figure is broken out by speaker group, care setting or language. There is no responsible AI statement, bias or fairness position, subgroup analysis, error taxonomy, model monitoring description or evaluation output. The trust center may hold a governance document, though none is public.

The terms of service address liability in both directions, and clinical risk sits with the customer.

Section 5.4, set in capitals rather than buried, states that Suki does not guarantee the accuracy or completeness of output from its artificial intelligence features and places responsibility for final review on the customer. It then excludes any liability whatsoever for the accuracy or completeness of processed data, or for any decision made or action taken in reliance on it. Processed data is defined to include clinical notes, orders, diagnoses and transcripts. A product that stages orders and assists with diagnostic coding therefore disclaims liability for the accuracy of exactly those outputs, and the clinician who signs owns the content entirely.

The indemnity running from Suki to the customer is narrow. Section 6.1 covers third party claims that the platform infringes a United States patent, copyright or trade secret, and states that this is Suki's sole and exclusive liability for any claim of any nature related to intellectual property. It does not reach clinical output.

Two provisions carry more weight. Section 5.2 warrants that the platform will operate as described in its documentation in all material respects, with correction of reproducible errors as the exclusive remedy, a stated performance obligation rather than a pure disclaimer. Section 7 caps direct damages at the fees paid in the twelve months before the claim, a defined ceiling rather than an open exclusion.

The patient has no route, no relationship and no standing.

Integration runs in both directions across five major EHRs, including Epic, Oracle Health, athenahealth and MEDITECH, and a partnership program extends to MEDENT, Azalea Health and WellSky. Generated notes are organized into LOINC coded sections, ready to be reviewed, edited and submitted into the EHR.

The voice command layer needs write access to ordering and staging workflows, a deeper integration surface than note insertion. Ambient order staging launched in April 2025 with prescription orders in athenahealth. A clinician speaks the recommended orders, and Suki structures, codes and stages them in the EHR for review and sign off. At launch, Suki named labs, imaging, referrals and other order types across Epic, Cerner and MEDITECH as the next step.

The athenahealth relationship runs deepest. athenahealth named Suki a preferred solution partner for ambient intelligence on 26 January 2026, a designation that brings fast track access to new athenaOne features. athenaOne customers can use Suki for Clinicians from athenahealth Marketplace, or Suki's technology embedded in athenahealth's own Ambient Notes. Through Suki for Partners, other healthcare software companies embed the same capability in their own products.

Depth inside the EHR is the center of Suki's positioning. The company describes deep real time integrations with Epic, Oracle Health, athenahealth and MEDITECH and calls itself the most embedded ambient AI solution on the market. The routes are named, including MEDITECH Expanse documentation APIs and athenahealth's Ambient Notes at general availability, which lets a practice run Suki inside a standard athenahealth interface.

A second route runs through other companies. The developer platform and the Suki Reseller Program let partners embed the core technology in their own products, so the software reaches clinicians under other brands. That widens reach, and it also means an end user may not know whose model is producing the note.

The app runs on desktop and on iOS and Android devices, a Zoom partnership covers telehealth, MedStar Health is named as a large multi site deployment, and continuous availability is marketed. Data at rest is stored in Google Cloud services such as Cloud SQL and Google Cloud Storage, and a Cloudflare web application firewall protects edge traffic. No region or residency commitment is stated for that storage.

Suki prices through enterprise channels, and its product structure is public, with two named tiers, so a buyer knows what is being priced before the demo and the enterprise contract that follow.

List prices, a free tier and a free trial are not offered, so a clinician cannot test the product on their own notes before contracting.

Coverage comes from reach. The homepage claims use across more than 100 specialties, on desktop and on iOS and Android devices, in various care settings. Integration with Epic, Oracle Health, athenahealth and MEDITECH puts the product inside the EHRs most US health systems run, from large integrated networks down to ambulatory practices. The MedStar Health deployment spans ambulatory specialties across more than 300 care locations.

A Zoom partnership brings it into telehealth encounters, and the developer platform carries it into third party applications in settings Suki does not sell into directly. On the developer platform, ambient sessions accept more than 80 languages, and the note always comes out in English for the EHR. Functionally it reaches past note writing, combining documentation, coding, clinical reasoning and question answering with voice commands for ordering and charting.

The specialties are counted rather than listed, and outcome figures are published only in aggregate.