Menaela

The staff assistant, the patient channel and the scribe all run on the model, while a practice management system with value of its own sits underneath. Main Menaela turns plain requests in chat or voice into actions such as finding a patient, checking availability or booking an appointment. The patient channel answers and books on WhatsApp, and AI Scribe drafts the clinical note from a recorded appointment.

Remove the model and scheduling, patient records, forms and the CRM still work through the regular screens, which the vendor says stay fully usable. The pitch is the combination: one plan positioned against the cost of a front desk person plus separate software.

Both sides of the boundary are published, along with the controls a clinic uses to change it. On WhatsApp the assistant may answer configured clinic facts and book, move, cancel or confirm appointments. It may not give patient specific advice on diagnosis, treatment, medication, risk, results, suitability or urgency; those go to staff, and emergency wording gets brief urgent care guidance plus escalation.

The stated rule for answering is whether the clinic has supplied enough authoritative information and the workflow permits it; otherwise the assistant says it cannot confirm and routes to staff. Ambiguous requests get a clarifying question before anything is committed, and failed tool calls are reported as failures.

The docs set out each override: pausing autonomy clinic wide, taking over one conversation until staff hand it back, a staff reply superseding a pending AI reply, and paused messages not being replayed later. Escalation alerts staff without stopping the assistant unless someone takes over. Spotting that a message is medical is still a model judgment, and how often that call is missed is not published.

A buyer can read how the system is built but not what it is built on. The security page explains that a language model acts as an orchestrator choosing among registered application tools, that tool arguments are validated against explicit contracts before anything executes, and that each clinic runs on its own PostgreSQL database. No model, model provider, speech recognition supplier, version or update practice is named, so a buyer cannot tell what produces a draft note or a patient reply, or when that changed.

The data path is described by category and no model provider is named. The privacy policy and data processing addendum authorize sub processors for hosting, communications channels and speech and language processing, under written terms with equivalent obligations, and say a current list is available on request with notice of material changes. WhatsApp is the one channel named. Who processes consultation audio, who generates notes and replies, and who hosts the clinic databases are not stated.

No named deployment and no measured result. The case studies page says detailed stories from pilot clinics are on the way and invites clinics to become pilots. The vendor deliberately publishes no savings figure, saying the economic case is clinic specific and should be modeled from the clinic's own workflow rather than a universal percentage. The homepage comparison of $3,000 for a secretary plus software against $399 for Menaela is a cost framing, not a performance result. The founder's record of launching more than 15 clinics is experience, not evidence about this product.

How information reaches the model is described in detail; retention and training use are not. The AI receives a fixed clinic context and works through permissioned application tools rather than querying the database. A patient conversation is bound to that patient and gets a narrower tool set, patient facing tools return approved care information rather than raw note text, and the raw scribe transcript is kept separately for review and audit.

No retention period is stated beyond the clinic's own configuration, nothing says whether clinic data is used to train or improve models, and de identification is not addressed. Safety engineering is published: the safety page walks through eight failure modes, including invented clinic facts, stale booking slots, the wrong patient context and tool failures, and how each is contained.

A substantive privacy framework reaches the service, and no business associate status is stated. The privacy policy and data processing addendum make the clinic the controller and Menaela the processor acting on documented instructions. They treat patient health data as a special category, list security measures, commit to helping with data subject requests, breach notification and impact assessments, and return or delete data at termination.

GDPR is named only where it applies, and the privacy law of the market the product is built for is not named. The homepage speaks of protected health information, yet HIPAA and business associate agreements are not mentioned, so a United States practice cannot tell whether one is offered.

Controls are described and nothing independent stands behind them. Published controls include a separate database per clinic, role based permissions applied to every AI action, encryption in transit, validation of model requests and an audit trail of agent actions recording actor, tool and target. No certification, external audit or penetration test is claimed, and there is no trust center. The vendor offers a walkthrough of tenancy, tool execution and audit behavior on request.

The product stays on the administrative side of the line, and its scope matches that position. The terms say AI output, including draft clinical notes, is for review by qualified staff and is not medical advice or a clinical decision, and that the clinician stays responsible for clinical judgment and for signing the record. Patient specific medical questions sit outside the autonomous patient toolset and are routed to staff. No regulator or regulatory regime is named, and the terms do not say whether the product is or is not a medical device.

Three languages are a selling point and nothing is published on how the system performs in each. The FAQ describes a Hebrew medical language pack and Israeli clinic terminology for transcription and note generation, rather than literal machine translation, alongside English and Russian. No accuracy figure, test method or comparison by language, accent or specialty is published for the scribe or the patient channel.

Governance of model behavior is described as architecture, with scoped tools, deterministic business rules and an audit trail, but no evaluation process, review material for a customer's own committee or external audit stands behind it.

Mechanisms let a clinic check what the AI did, and nothing stands behind its output on the published terms. Draft notes stay drafts until a clinician signs, signed notes are versioned and protected from regeneration, the raw transcript is kept for review, bookings are validated by the scheduling layer before they are written, and agent actions are logged. The terms provide the service as is, place liability limits in each order and make the clinician responsible for clinical judgment. No accuracy commitment, remediation obligation or indemnity is published, and the terms name neither the contracting entity nor a governing law.

Menaela replaces the practice record rather than connecting to one, and no clinical system integration is published. WhatsApp is the live patient channel, with its setup documented. The integrations page labels website lead endpoints, email lead intake and Meta Lead Ads as roadmap items not available today, though the FAQ describes the same lead sources in the present tense. No electronic health record, laboratory, prescribing or billing connection, data export route or interface standard is named, and migration from an existing system is a scoped onboarding item.

One hosted service, with tenant isolation documented and location left open. Each clinic gets its own PostgreSQL operational database on shared application infrastructure, and the AI cannot switch to another clinic's data. The homepage says residency, processors, transfers and retention have to match the clinic's jurisdiction, and cross border transfers rely on standard contractual clauses, but no hosting region, cloud provider or residency commitment is stated.

A price and a unit are published, and the route in is a conversation. The pricing page lists a Core plan at $399 a month for solo practitioners and small clinics, covering practice management, scheduling, patient communications, the staff assistant, AI Scribe and the CRM. It also names what changes the scope: migration from an existing system, custom integrations or channels, multi location setups and custom workflows. Onboarding is scoped per clinic and not priced. There is one plan rather than tiers, every call to action leads to a contact form, and the FAQ still tells readers to contact the team for current pricing.

Coverage is stated clearly and nothing validates it yet. The product is built for solo practitioners, small private clinics and clinic owners, in English, Hebrew and Russian, with a right to left Hebrew interface, a Hebrew medical language pack for transcription and notes, and Israeli clinic terminology. No specialty is named; the sample note on the homepage is an outpatient knee pain visit. Multi location clinics are listed as a scope item rather than a standard setup. No live clinic is named in any setting, country or specialty.