Weave

An established communications and payments business that has put a model driven receptionist at the center of its product direction. The underlying asset is the platform: practice phones, texting, reminders, scheduling, digital forms, payments, reviews and email marketing, with integrations into practice management systems. The AI Receptionist answers calls, texts and website chats around the clock, books and reschedules appointments for qualified dental practice management systems, captures requests for staff elsewhere, and hands over to staff; the company's May 2026 announcement says it is built with Google Cloud's Gemini Enterprise Agent Platform. Call Intelligence applies AI to call recordings.

The phones, messaging and payments would continue to function with every model removed, and the reasoning engine is Google's rather than the company's own. Still, the agent does substantive work, answering and booking rather than decorating a workflow.

No figure shows what share of calls the receptionist resolves without staff, and nothing explains the healthcare specific layer above Gemini.

The receptionist answers on its own, hands over when unsure, and the practice chooses when it takes calls. The AI Receptionist page says practices select when calls route to the receptionist; that when it meets something it cannot answer confidently it routes the call or message to the team with a conversation summary; that appointment requests are captured for staff to review and schedule, with automatic booking and rescheduling only for qualified dental practice management systems; and that every conversation, human or AI, sits in one inbox. Context carries across calls and texts so a patient is not asked to repeat themselves.

What is missing is the confidence threshold and the after hours destination. The page sells answering calls day or night, and its illustrated night shows the receptionist booking an emergency appointment at 6:05 am, but nothing describes how a caller describing pain, swelling or bleeding is handled when the team it would route to is not there, or whether callers are told they are speaking to an AI.

The model supplier is named and the version discipline is not. The company's May 2026 announcement says the AI Receptionist is built with Google Cloud's Gemini Enterprise Agent Platform and leverages Google Gemini, and the AI Receptionist page says it uses advanced conversational models trained on real world customer interactions, preserving context across voice, text and chat and routing to staff when not confident.

What is not published: which Gemini models are used, what the healthcare specific layer above them contains (prompts, rules, fine tuning or integrations, which fail differently), what the customer interaction training data is, any accuracy, containment or transfer figure, and how model changes are tested and announced to practices.

The model provider and the host are both named. The company's May 2026 announcement says the AI Receptionist is built with Google Cloud's Gemini Enterprise Agent Platform and leverages Google Gemini, and the security page says the platform runs on Google Cloud, including Cloud SQL and Google's key management service. So the party whose models process patient calls and texts is identified, and a buyer can read Google's own enterprise terms on retention and training and ask which of them this configuration uses.

What is missing is the rest of the list and the terms that bind it. No subprocessor list is published, telephony and payment processors go unnamed, and nothing states whether conversation content is excluded from Google's training, where it is retained, or whether the business associate agreement reaches Google's processing.

Large, verifiable scale and named customer results, with no measurement of the AI. The company reports roughly 40,000 customer locations in its securities filings, which makes the figure more reliable than a private vendor's claim. The pricing page cites Smith Dental's 28 percent increase in new patients and Beaches Dental saving $3,500 a year, and the AI Receptionist page states that one in three calls is missed, 80 percent of missed calls are booking requests and 47 percent of bookings happen after hours, which frames the problem rather than measuring the product.

No published figure measures the receptionist: no resolution or containment rate, no booking accuracy, no transfer rate, and no comparison with the staff it replaces. None of the customer figures states a period or method.

Security assurances are published and the questions the receptionist raises are not answered. The security page describes encryption in transit and at rest with keys managed through Google's key service, and says the company does not use or disclose personal information except as its privacy policy and subscriber agreements permit. The AI Receptionist page says the receptionist is designed to meet HIPAA standards and that its conversational models were trained on real world customer interactions.

That last sentence makes the training question concrete: nothing states whose interactions trained the models, whether practices' patient calls and texts are used to train or improve them now, whether Google's platform retains any of it, or how long call recordings, transcripts and messages are kept. Nor is any opt out for practices described.

HIPAA compliance is claimed, and no document on the security or receptionist pages backs it. The security page says the platform has been designed with features to support practices in complying with HIPAA and links a help article on HIPAA compliance; the AI Receptionist page says the receptionist has been designed to meet and exceed HIPAA standards. Neither page states business associate status or offers an agreement, though the help article and the privacy policy may say more.

The subprocessor point is concrete here: the receptionist is built on Google's Gemini Enterprise Agent Platform and the platform runs on Google Cloud, so a second organization processes patient calls and messages, and a practice needs the agreement to reach it.

Controls are described with outside checks behind them, and no attestation of the company's own. The security page says independent third parties conduct regular penetration tests of the websites, web applications, interfaces and cloud infrastructure, that systems are scanned for vulnerabilities, and that a bug bounty program follows Bugcrowd's vulnerability rating taxonomy. It also describes encryption, multi factor authentication, annual security training with phishing exercises, an incident response plan and vendor risk management.

The only SOC 2 and ISO 27001 reports mentioned are Google Cloud's, which cover the host rather than the company. No SOC 2 report, HITRUST certification or trust center of the company's own is published, and the penetration test cadence and firm are not named.

No clearance, authorization or submission is claimed, and none is needed. Scheduling, reminders, insurance capture and payment collection are administrative. Three other regulatory frames apply and none is device law: automated outbound patient contact falls under telephone consumer protection rules; payment processing brings card industry security requirements; and insurance eligibility verification operates inside payer contracting. Assessing this product against a device framework means examining the wrong law.

Nothing published on how the receptionist's behavior is governed or tested. There is no evaluation method, subgroup performance, model documentation or drift monitoring, and the company's July 2026 statement that enterprise customers want more visibility into AI performance describes demand rather than a published answer.

The exposure is the caller population. A receptionist answering practice phones and texts around the clock meets every accent, dialect and first language its customers' patients speak, and a caller it understands less well waits longer, is routed less accurately or gives up. The AI Receptionist page lists real time language identification and switching as coming soon, so today's behavior with callers who do not speak English is not described.

Mechanisms exist to see and check what the receptionist did, and nothing stands behind it. Every conversation, human or AI, lands in one inbox; a handoff carries a conversation summary; appointment requests outside qualified dental systems wait for staff to review and schedule; and Call Intelligence lets the team analyze calls. A practice can therefore find and correct a wrong booking or a mishandled call.

No accuracy or containment figure, warranty, service level on the receptionist, or remediation commitment is published, though the terms of service may say more. The receptionist books and reschedules directly into the calendar for qualified dental systems, and a wrong booking or a missed urgent call surfaces only when staff read the inbox.

Operational integrations with practice management systems, as the company describes them. The pricing page lists integrations for dental, medical, optometry and veterinary practice software, and the AI Receptionist page says it books and reschedules appointments in real time into the calendar for qualified dental practice management systems and supported integrated systems, and captures requests for staff elsewhere. The company describes its integrations as driving scheduling, insurance verification and payment collection across roughly 40,000 locations, in a fragmented market where each connection is separately built and maintained.

The pricing and receptionist pages name no system, the named list sits on a separate integrations page, and no interface standard or implementation detail is published. Nothing separates systems with two way booking from read only ones, or says how the receptionist behaves when a connection fails.

A single hosted service on a named cloud, with the region implied rather than committed. The security page says the platform runs on Google Cloud, with Cloud SQL backups at least daily, high availability across zones with synchronous replicas and automatic failover, Kubernetes deployments, and the ability to redeploy a full copy of production in a different Google Cloud region after a region wide outage. That is a clear continuity description.

No hosting region or residency commitment is stated, and nothing distinguishes what is retained from what is processed and discarded. The platform is both a phone system and a messaging system, so it holds the content of patient conversations, and the receptionist adds processing on Google's agent platform. Whether a regional failover could move data outside the United States is not addressed.

A starting price is published and the tiers are laid out without figures. The pricing page says plans start from $199 a month and sets out three tiers, Pro, Elite and Ultimate, with a feature comparison including monthly bulk text allotments of 1,500, 3,000 and 15,000 messages and up to 15 phones on Ultimate; each tier resolves to a pricing request, with enterprise pricing separate. Customer support and onboarding are included in every plan. The AI Receptionist is not priced on the plans page and is sold through an early access sign up. The contract term is not stated.

The company is listed on the New York Stock Exchange, so its financial position is disclosed in its filings; that tells a buyer about supplier durability rather than about price.

Wide within a chosen band of small and medium practices. The customer base is concentrated in dental and optometry, with medical and veterinary practices and a growing multi location segment, across roughly 40,000 locations, and the plan comparison carries dental only, optometry only and veterinary only features. Functionally it spans the front office: phones, texting, scheduling, forms, insurance verification, payments, reviews, email marketing and now the AI receptionist.

The named results come from dental practices (Smith Dental, Beaches Dental). The receptionist is not yet available to veterinary offices, and automatic booking works only with qualified dental systems. Veterinary practices are part of the location count and are not human healthcare. Geography is the United States.