Cost Math: Flat Rate Versus Enterprise Quote
Suki AI does not publish a self-serve price. Contracts are quote-based and usually negotiated through health systems or EHR partners. Publicly reported figures generally fall between $299 and $399 per provider per month. Merry AI Pro Annual is $54 per month, or $648 per year, with no per-encounter metering.
Run the arithmetic per clinician. At $299 per month, Suki totals $3,588 annually. At $399, it totals $4,788. Against Merry, that is $2,940 to $4,140 saved per physician, an 82% to 86% reduction. A six-physician primary care group keeps $17,640 to $24,840 a year before any implementation fees. For market context, Freed lists $99 per month, and Nuance DAX Copilot enterprise contracts commonly run near $8,000 per year.
Cost alone should not decide this. Suki's native, bidirectional integrations with Epic, Oracle Health, athenahealth, MEDITECH, and eClinicalWorks are real engineering assets. They suit health systems that have integration teams. The question for an independent practice is whether that architecture is worth five to seven times the price.
| Dimension | Merry AI | Suki AI |
|---|---|---|
| Pricing | $54/mo Pro Annual, flat | Quote-based; reported $299–$399/mo per provider |
| EHR write path | Chrome extension DOM injection into visible fields | Native vendor integrations and SDKs |
| DOM injection speed | Sub-second field write after draft generation | Not applicable; sync timing depends on connector |
| Group note-splitting | Diarized per-patient drafts from one session | No documented multi-patient splitting |
| CPT G2211 capture | Flags qualifying visits; suppresses on modifier 25 conflicts | ICD-10 suggestions; no published G2211 logic |
| Attestation logging | Clinical Logic Bridge sentence-level source log | EHR-native audit trail of note writes |
| Data retention | Audio discarded after note finalization | Audio and transcript deleted at 30 days; note kept for contract term |
Workflow Friction in eClinicalWorks v12
How Field Detection Works on the Progress Note
When you open an eClinicalWorks v12 progress note, the Merry extension reads the page structure. It identifies the HPI and Assessment fields by label and container, then confirms which tab is active before writing anything. It inserts the draft and dispatches standard input events, so the EHR treats the text as typed. Nothing is saved until you review and sign.
Suki takes a different path. Its documentation describes notes flowing back through a native integration into the proper chart sections, with assessments mapped to the problem list. An EHR vendor can govern that approach more easily. Its limitation is coverage. Only fields the connector exposes are writable, and configuration depends on your instance and contract.
Candidly, the tradeoff is real. DOM injection works on whatever is on screen, but it bypasses server-side validation. Merry compensates in three ways. It never writes into billing or order fields. It restricts insertion to narrative sections the clinician must sign. It logs each insertion. Selectors are revalidated after every eClinicalWorks release.
Group Visits and Note-Splitting
Shared medical appointments are common in primary care for diabetes, hypertension, and heart failure education. Merry diarizes speakers in a single session and produces a separate draft for each enrolled patient, attributing statements to the correct chart. Suki's public materials describe single-encounter ambient sessions. We found no documented multi-patient splitting.
Heart Failure and G2211 Coding Accuracy
Capturing LVEF and NYHA Class During E/M
Whenever heart failure management is discussed during an E/M visit, Merry extracts the stated LVEF percentage and NYHA functional class and places both in the Assessment. The distinction matters. An LVEF of 40% or less supports HFrEF coding, 41% to 49% indicates mildly reduced ejection fraction, and 50% or above indicates HFpEF. Each maps to different ICD-10 I50 subcodes and different guideline-directed therapy expectations.
Accuracy here means provenance, not inference. If the clinician cites a March echo showing 35%, Merry records the value with its stated source and date. If LVEF or NYHA class is never spoken, the field stays blank and the clinician sees a prompt. Merry does not estimate either value from symptoms alone.
G2211 is a modest add-on with meaningful aggregate value. CMS permits it with office and outpatient E/M codes 99202 to 99215 in two situations. The first is when the practitioner serves as the continuing focal point for a patient's care. The second is ongoing care for a single serious or complex condition. National payment sits near $16 per claim.
The recovery math is straightforward. Four qualifying Medicare visits per clinic day across roughly 245 days produce about 980 claims, or $15,600 or more per physician each year. Merry also suppresses the suggestion when modifier 25 is appended. CMS added three exceptions for 2025: an annual wellness visit, vaccine administration, or a Part B preventive service on the same day. Merry still suggests G2211 in those cases.
Audit Defense and Data Retention
Every inserted sentence carries a source record. The Clinical Logic Bridge links each Assessment statement to its transcript excerpt. The attestation block stores clinician identity, timestamp, and the longitudinal-relationship rationale supporting G2211. When a payer requests records, the practice can show why the code was billed, not only that it was.
Retention policies differ in detail. Suki's developer documentation states that ambient audio and transcripts are permanently deleted after 30 days. The clinical note is retained for the duration of the service contract. Merry discards audio once the note is finalized, and the note of record lives in your EHR.
For the underlying peer-reviewed literature on ambient documentation and clinician time, search https://www.ncbi.nlm.nih.gov/pmc/. If you would like to compare your current workflow against these numbers with a colleague, Schedule a 15-Minute Workflow Audit.
