Head-to-Head Architecture Comparison
When solo and small-group internists evaluate ambient documentation, the real question is rarely note formatting quality, since both Merry AI and Suki AI produce structured SOAP and H&P drafts that read well. The decision comes down to three quieter variables: how text actually arrives in the chart, how much the software costs across a full year of high-volume encounters, and whether longitudinal complexity is captured at the point of care or reconstructed later. This review works through each the way a colleague would review a chart with you, without slogans.
Suki AI's published integration model follows a four-phase pattern: authenticate for a session token, open an ambient session tied to an encounter, stream audio over WebSocket to the backend, then retrieve the generated draft through REST endpoints for review and EHR handoff. This is a sound enterprise design. It also introduces a structural dependency on API queues and, in many deployments, an integration project with your EHR vendor before the first note is written.
Merry AI takes a deliberately different path. It runs as a Chrome extension that captures the room conversation, drafts the note locally against your chosen template, and injects the finished text directly into the EHR field where your cursor already sits. There is no session-token handshake for the clinician to manage and no backend retrieval step between session close and text placement. The architectural consequence is that the two systems differ most where clinicians feel it: the seconds between finishing a visit and having a signed note.
| Feature | Merry AI | Suki AI |
|---|---|---|
| Monthly cost (annual billing) | $54/mo Pro | $99/mo standard |
| DOM injection speed | Direct tab injection, sub-second | API retrieval then paste |
| Multi-speaker group note-splitting | Speaker diarization, labeled segments | Single-author draft |
| CPT G2211 complexity prompting | Real-time visit-close prompt | Manual add-on |
| Attestation logs | Cures Act timestamped, edit history | Basic text export |
| Data retention control | Configurable, BAA-covered | Enterprise-governed backend |
| EHR deployment | Zero-IT browser extension | Integration project |
The Real Cost of Metered Documentation
Pricing is the section clinicians most often skip and most often regret skipping. Suki is sold as a per-user subscription in the range of $99 per month for standard access, with ambient processing governed by contract terms. Merry AI's Pro plan is $54 per month on annual billing with no per-encounter metering. Across a single clinician year, that difference is roughly a 45% reduction in software overhead before any coding recovery is considered.
The metering question matters more in internal medicine than in most specialties, because panel-based ambulatory care produces 20 to 30 encounters daily, every one of them a candidate for ambient generation. When ambient sessions are treated as a monetized unit, the incentive structure quietly pushes clinicians toward shorter sessions or selective use, which undermines the reason to adopt ambient documentation at all. Un-throttled generation removes that tension. You can review the full plan structure at Compare Practice Partner Plans.
A predictable flat rate also simplifies practice budgeting. There is no reconciliation of overage fees at the end of a busy quarter and no negotiation over usage tiers when you add a new advanced practice provider. For a two-physician ambulatory clinic, the annual delta between a $54 seat and a $99 seat is roughly $1,080 saved before the coding math even begins.
Zero-IT Deployment Versus Integration Projects
Deployment timeline is a cost that rarely appears on the invoice but always appears on the calendar. Suki's model, integrating through EHR APIs such as Epic or MEDITECH endpoints, typically involves an implementation phase, identity and access mapping, and validation before clinicians go live. That is appropriate for a large health system with an IT department. It is a genuine barrier for a small internal medicine practice.
Merry AI's browser-extension architecture requires no server-side integration project. A clinician installs the extension, authenticates, selects a template, and drafts a live note the same day. The DOM injection engine works against the EHR's existing web interface, so there is no dependency on a vendor API roadmap or a scheduled integration window.
Coding Accuracy and G2211 Capture
The largest recoverable dollar figure in ambulatory internal medicine right now is the G2211 visit complexity add-on. CMS introduced it to reflect the inherent complexity of longitudinal, relationship-based care, and per AMA CPT guidance at https://www.ama-assn.org/practice-management/cpt it reimburses approximately $16 per qualifying visit. The trouble is that it is chronically under-captured because it depends on documenting a longitudinal reasoning pattern, not a single billable procedure.
Suki's ambient order staging extracts medications, labs, and imaging with appropriate coding, which is useful, but its published capabilities do not include automated G2211 identification. The add-on remains a manual decision the clinician must remember to make. Merry AI surfaces a visit-close prompt when longitudinal-complexity indicators appear in the transcript, so the code is considered while the clinical reasoning is still in working memory rather than reconstructed during month-end billing review.
Attestation Logs and Audit Defensibility
Documentation that recovers revenue is only valuable if it survives a payer audit. Merry AI writes a Cures Act compliant attestation block with every finalized note, recording the generation timestamp, the reviewing clinician, and the edit history distinguishing proposed text from accepted text. This establishes human-in-the-loop authorship on the record rather than as an assumption.
The Clinical Logic Bridge attaches the specific complexity indicators supporting any G2211 attestation, so an auditor reads the documented rationale alongside the code. Suki's outputs support review and export, but attestation timestamping and complexity-rationale linkage are not advertised as core guarantees, which leaves the audit trail to be assembled after the fact.
Compliance, Retention, and the Human-in-the-Loop
Both platforms operate under HIPAA as business associates and require a signed BAA before any protected health information is processed. Merry AI's BAA covers the ambient capture, local drafting, and DOM injection pipeline, and data retention is configurable so a practice can set retention windows that match its own governance policy rather than accept a fixed backend default.
DOM injection security deserves a direct answer, because writing into an EHR field sounds invasive until you see the mechanism. The extension does not scrape or transmit chart contents; it places the clinician's own drafted note into the field the clinician has focused, using standard browser input events. The finished note is authored by the clinician who reviews and accepts it, and nothing is written without that explicit action. This preserves the same human-in-the-loop control point that Suki enforces through its review-and-approve step.
Which System Fits an Ambulatory Panel
For a large integrated health system with a dedicated IT function and an existing Epic ambient program, Suki's enterprise integration model is a reasonable fit. For an independent internal medicine practice measuring cost per clinician, deployment friction, and G2211 recovery, the calculus favors a flat-rate, zero-IT extension that captures complexity at the point of care. If you want to walk through your own numbers with an engineer, Book a 15-Minute Workflow Audit, and to see the template library your practice would inherit, Access 1,000+ Specialty Clinical Prompts.
